Paul Bryant's Networking Correspondence
Present: J Barlow, P Bryant, T Daniels, J Hutton
Approved.
P Bryant had reminded the PITs list that it should not be used for discussions. Unfortunately a certain amount of discussion continues although the level is not high.
The VMNET links to Europe were suffering from poor performance. This manifested itself by links timing out. However, the queues were at a fairly low level. The EBONE connections should improve matters.
The EARN regional sites are being encouraged to be "INTERBIT gateways". This entails sending mail with "good" domain addresses via SMTP. This can be done in two ways. The first entails re-routing RSCS traffic to SMTP which was destined for a remote gateway. The second is to amend the MAILER tables to directly route appropriate traffic to SMTP. The second method is proposed as this will allow better control and also a slightly larger range of traffic.
This is approved but it should be done in stages with careful monitoring of the effect.
The request to use VMNET to Italy and France was approved although it should if possible e delayed until EBONE is in place. The VMNET connection to Germany was approved.
The EARN BOD will meet on May 10/11. The Executive is proposing a new funding model for 1993 which substantially reduces the UK contribution. It is proposed to apply the model retrospectively to 1992. J Hutton thought that the model would be acceptable to the UK.
July 21, 1992 at 9am.
Members: P Bryant, G Christian, P Girard, T Kidd, R Westlake, J Wheeler
1 Decide technical objectives
2 Decide programme of work
3 Time scale
4 Next meeting
Present: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, D Parker, R Westlake, J Wheeler
To plan for possible elimination of the IBM as a provider of critical mail services such as gateways.
To re-consider the mail requirements for work stations including PCs.
To plan for the use of multiple mail protocols including RFC822, Grey Book and X.400.
As a result of the above activities, to produce a new mail strategy and an implementation plan.
Mount PP on DIRECTORY initially and an alternative machine later and evaluate its use. Mounting should be complete by August 17. Action R Westlake
Make case for the purchase of a Spark station for development use and in particular as a better platform for evaluating PP. The case should go to the August 3 TMPC. Action K Hoadley
Evaluate possible move of the EARN gateway to use PP. Action G Christian
Examine the use of NJE under UNIX. Action T Kidd
Study possible mail user interfaces such as SMAIL. Action P Cahill
Study alternatives to LISTSERV such as MAILBASE and INFOSERVER. Requests for information will be made via INFONETS and UKMAIL lists. Action J Wheeler
Produce draft strategy. Action P Bryant
Study the problem of mail originating in X.400 for delivery within EARN. Action ?
Study addressing styles within PP. Action ?
Friday July 31, 9am, Colloquium.
Mail services have been centralised to some extent on the IBM 3090 with the use of MAILER/EARNGATE with its name look up facilities and LISTSERV/LISTRAL providing mail distribution. SMTP and Grey Book mail is provided for use to local and remote machines. SMTP allows the use of the IBM as a mail server for some workstations.
The IBM provides a gateway to EARN which uses RFC822 over BSMTP.
The policy has been for the IBM service to accept mail for anyone on the RAL site and forward it to the appropriate machine.
Recently circumstances have changed. X.400 services are expected to be required as a result of the JNT initiative, mail services to workstations need further development, attractive mail products such as PP and MAILBASE have become available and in addition the long term future of the IBM 3090 for mail purposes is in doubt. This suggests that a review of mail services is needed with a view to moving them to alternative platforms.
The mail service should follow a number of principles:-
There are four requirements - mail transfer agent(s), message store(s), mail user agents and a mail distribution service.
Although there may be a number of mail transfer agents there must be a central one to which all incoming mail will be directed. This must have a name look up facility with a directory of all people on the RAL site.
A mail transfer agent(s) is required to act as a gateway between Grey Book mail, BSMTP (EARN), SMTP and X.400.
The principal contender all the above requirements is PP which is provided and promoted by the JNT and requires a UNIX platform.
A message store can also be provided by PP. There may be other alternatives.
Mail user agents are required for each machine. In the short term it is a matter of expediency as to which protocol is used. In the longer term SMTP and/or X.400 agents are expected to predominate.
There is no obvious choice for the mail distribution list server. Further study is required to identify a suitable server.
There is an advantage in providing services on UNIX platforms. This gives the best protection for the investment and also allows a wide choice in the selection of hardware and its subsequent upgrading.
The prime candidate for a mail server is PP. This runs on a SUN Spark station. Indications are that PP can provide the central mail transfer agent, a gateway function to all protocols apart from EARN mail and a message store.
Mail transfer agents are available for most protocols on most platforms.
Currently products are available for implementing the proposed strategy. The quality of the products and their technical characteristics is not clear.
The JNT is expecting a rapid migration to X.400 towards the end of 1992. This transition is expected to be slower than the JNT expect and may not result in a single ubiquitous mail protocol. Thus a measure of caution is suggested to keep options open.
PP should be mounted and evaluated as a central mail transfer agent and gateway.
The preferred mail protocol stack(s) for use on the local area network should be identified.
user agents on popular machines implementing the protocol stack(s) should be evaluated and recommendations made.
Mail distribution services should be evaluated and recommendations made.
Further recommendations depend on the results of the evaluations and studies and expected to lead to purchases of hardware and software to provide services.
Present: J Barlow, P Bryant, T Daniels, J Hutton
Regretfully these were not circulated.
There have been no significant performance problems. The queues have been at a low level.
There has been a modest increase in traffic which is probably more to do with seasonal fluctuations.
A VMNET link has now been set up to CERN with the approval of HEP as the connection uses their line.
During this month mail traffic with "good" domain addresses has been directed to SMTP in line with EARN/BITNET policy. Initial indications are that this will remove rather more mail from EARN than was expected. It was agreed to provide SMTP statistics to J Hutton. These statistics will include traffic to local workstations, to the rest of the UK and pre-existing SMTP traffic so will only be a rough guide to the traffic which has moved.
At the end of September the maintenance of the G-BOX ends. there are three options:
Germany will be pressed to set up a VMNET link during August.
It is conformed that JNT may have the G-BOX machine at no cost when it is no longer required for EARN purposes.
It is understood that Alister Chalmers is drafting a letter to F Greisen to agree to the EARN subscription proposals.
JNT would now be happy to accept a properly signed and dated invoice from EARN and P Bryant will inform EARN. Action P Bryant
P Bryant agreed to determine whether the UK had one or two sites (RAL and ECMWF). (It is now confirmed that the UK has one site).
September 29, 1992 at 9am.
The strategy detailed in this paper proposes a transition of parts of the local area network to use 100 Mbit/sec FDDI technology. The elements of the proposal are:
Our ethernet has been with us for four years and has been a great success. It is based on a backbone ethernet bridged to village ethernets.
When the network was designed routers were expensive and had poor performance thus a bridged network was set up. If we were to build the network now we would probably use brouters to allow much better control over the traffic and better insulation between the parts of the network.
Parts of the current LAN are experiencing heavy loads for short periods and this is leading to some ad hoc additions of bits of subnetwork.
If we are not careful our network will grow in a rather messy way, become complicated and therefore may become less reliable.
We can expect a demand for higher speeds although at this stage it is difficult to predict the magnitude of this demand. The imminent expectation of "SUPERJANET" with speeds well in excess of the current 10 Mbit/sec ethernet is likely to trigger a demand to exploit this service from various parts of RAL and also to services within RAL.
Any new network must be a graceful development from the current one with no demand for expenditure except for those wanting better facilities.
Any new network must encompass ethernet, FDDI and possibly other technologies.
The network must support our favourite protocols - IP, Pink Book, Novell, DECNET phase IV, DECNET phase V and its friend CLNS. It is likely that other protocols will need to be carried.
Good management facilities are required.
There are currently 13 villages bridged onto the backbone. Several villages have further bridges for convenience or to reduce the load on some segments.
The backbone is not overloaded as most traffic is within a village although there is occasional saturation for short periods.
JIPS, the IP connection to CERN and SUPERJANET are likely to increase the traffic load.
A study of technology indicates a number of constraints that seem unlikely to change in the near future:
A number of suppliers have been contacted and their equipment examined to satisfy us that the strategy is feasible.
A number of villages have shown interest in FDDI services which indicates that at least four will participate in the initial stages.
Fibre optic cables are provided through the established channels as there are no special considerations.
A provision timetable, subject to detailed discussions with suppliers, is:-
It is essential that all equipment is delivered and paid for in the 1992/3 financial year. In effect this means equipment acceptance by March 1, 1993.
It is recommended that the strategy in this document is adopted.
It is recommended that an initial installation covering at least four villages is undertaken after further equipment evaluation and selection.
It is believed that the above recommendation can be implemented within the financial year 1992/3.
CCD has a small FDDI network to connect the CRAY and IBM 3090 to the current ethernet network via a router. There is a concentrator which will be used to connect an IBM RS6000 and a graphics system. Apart from the equipment connected to the concentrator all equipment has been provided by NSC.
Our experience of FDDI, when delivered and software made available, is that it works well and reliably but equipment deliveries have been subject to delays.
To: Anne Barton From: Paul Bryant Date: 23 July 1992 Subject: Proposed tender
We intend to go to tender for about £100K of network equipment towards the end of August. Attached are two documents giving details.
The first document is the result of a working party's deliberations. The second is a draft for the tender exercise.
I would like suitable versions of both documents to go to possible suppliers. We may need to take out some text which relates to costs and refers to manufacturers. At this stage both documents are draft.
The possible suppliers are:
DEC, 3COM, NSC, Chernikeef, Wellfleet, Proteon, Cabletron, Datarange, Computer International, IBM.
Results of our evaluations so far suggest that 3COM will be cheapest but that DEC will give us greatest confidence that their equipment will work. NSC and Chernikeef will probably be too expensive. Wellfleet and Proteon are unknown quantities. Cabletron, Datarange and Computer International probably cannot meet the specification but are their as we have had discussions with them. IBM is included as a recent announcement suggests they may have something to offer.
At this stage I would value your help in preparing for the tender.
To: Division Heads From: CCD Date: July 27, 1992 Subject: A new local area network strategy
Recently an inter-divisional working party has produced a new local area network strategy. This was triggered by a certain amount of network congestion, the expected improved JANET bandwidth with SUPER-JANET and the performance demands from workstations. £100,000 has been allocated centrally for the introduction of an improved local area network to be provided this financial year.
The working party report is appended. It proposes the use of FDDI. The funding provided would cover the cost of connecting up to four "villages" or groups of users. If more than four wish to connect we propose raising the additional funds by way of a levy on each connected village. This will range from zero (4 villages) to £15,000 (8 villages). Although providing FDDI to the villages, any end user equipment for connecting to the network would be provided by the village.
It is proposed to go to tender for the required equipment towards the end of August by which time a definitive list of villages is needed.
Conditions for joining the network at a later date are not yet defined. It is expected that the cost of FDDI equipment will drop over the next few years.
May we ask for a statement of your intention to take part in this development?
Present: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
It seems Spark should be spelt Sparc.
Still expected mid August before the next meeting.
A proposal will go to TMPC in two weeks. It will be circulated. Action P Bryant
There is some inconclusive debase as to whether the EARN gateway should be on the same machine as PP and whether it should be logically separate.
HUJI had been obtained and mounted on the RS6000 but not yet run. It suffers from being an old version of NJE without multiple streams which is a serious draw back. UREP would seem a better bet but may cost money. Details of obtaining UREP are being obtained.
SMAIL is an MTA not a UA. It looks to be flexible and useful for workstations. It can drive both STMP and Coloured Books and may be able to act as a gateway. Cambridge added the Coloured Book part.
PINE V3 has been looked at. It is screen based and deals with multi media. There is a UNIX and PC version. A description will be circulated. Action J Wheeler
K O'Neill is looking at CC and talking to S Weston about White Book mail for PCs. White mail works over Pink Book. It is understood an RFC1006 version is being developed but will not be free as it is not sponsored by the JNT. CC seems to have the problem of needing gateways to SMTP and X.400 and their availability is unclear.
Some information has been obtained from a call for information. PP has some distribution list facilities. Apparently there is a LISTSERV for UNIX but details are not to hand.
An initial draft has been circulated. Comments are welcome on the first few sections but the last ones will be rewritten as the study progresses.
This is waiting for progress on mounting PP. Apparently X.400 is now available on the IBM as a product. At this stage there is no plan to look at it.
Waiting for PP to be mounted.
August 20
The beta test of White Mail is available as of today. K Hoadly will obtain it. Action K Hoadley
Over the last few months, the importance of network services within Rutherford has increased dramatically. In the near future the likely introduction of X400 and Kerberos will merely reinforce this trend. Though many of these services are barely visible to the end user, they are often essential to support the applications like electronic mail we all take for granted. Whilst in the past network services have been provided using whatever hardware was available at the time (the "beg, borrow or steal" approach), as the number of services grows and their importance increases this is unlikely to be satisfactory, and higher priority will need to be given to providing suitable hardware dedicated to support these services.
As Rutherford moves away from reliance on directly connected mainframe terminals and towards a more distributed computing environment, so the importance of the local area network increases. Though applications such as electronic mail are already in wide use on the local network, other new facilities, such as directory services and secure networked authentication, will be needed to support a user base which is growing in both number and sophistication. Over the last year a number of new services have started to come into use. However without a clear plan for their introduction they have been spread over a wide range of hardware, wherever sufficient resources could be found; a case of 'beg, borrow or steal'. Inevitably this has led to some less than optimal solutions; in particular the tendency to look towards ARCU for the provision of all new services. This situation would be of little consequence if the network services in question were of marginal importance, merely frills around the edge of the network. But this is not the case. The best example is the DNS, which though only recently introduced into the lab, has already become essential. Many of the other new services just starting to appear are likely to follow the same route. Thus network users depend on an ever increasing number of networking services, though often without being aware of the fact. Despite this dependency, these services enjoy at best a precarious existence running on hardware which is either overloaded or earmarked for other purposes. As the number of networked systems increases it is clear that the ad hoc means by which these services are currently provided will not suffice.
Over the next 18 months it seems certain that the number of different networking support services in use will increase. The list of services is likely to include:
Very few of these services are available on, or could be ported to the IBM. Though with some (probably considerable) effort most of these services could be ported to run on the Unix front-end, this would seem to be a distraction from that machine's primary purpose. Thomson, though perfectly adequate as a X500 and DNS server, would have difficulty coping as an X400 gateway in addition.
To meet this lack of suitable hardware, I propose that:
In view of the pressure from the JNT to bring PP into service this calendar year, and particularly the need for a development machine to experiment with PP beforehand without affecting production services such as the DNS, the need for a new machine is becoming urgent.
Note in the longer term, if the use of X400 takes off then it would be sensible to separate the mail gateway from the message store by providing one or more dedicated message store machines.
As more and more people start to use the local network, the quality of networked applications and support services becomes paramount. Given the growing number and importance of such services, and the lack of suitable hard ware currently present with which to host these services, there is a need for one or more dedicated network service machines.
Though there are numerous vendors of Unix workstations which might be suitable for a network services machine, it seems sensible to stick with those machines with which we have some experience. Thus there are three alternatives
In summary: It is proposed that a Sun workstation, together with X25 card and software, be purchased for the network services machine.
The cost:
SPARCserver II: 10300
monitorless, 32MB memory, 424MB SCSI disk
X25 High-speed Serial Interface (HSI/S) 2000
Sunlink X25 software 1724
-----
14024
-35% discount 4908
-----
9116
+17.5% VAT 1595
=====
10711
The current charging arrangements take no account of EARN traffic that uses gateways - thus:-
Both these subscription avoidance schemes cause a higher charge and possible unfair increase in charges for other countries.
The charging situation is becoming ever more complicated. For example - SMTP traffic is free at present but in effect the costs of this traffic come from countries paying for lines and also from donations to the Ripe Operational Unit. EARN is one of the benefactors of the Unit.
Charging can have an affect on traffic. With the scheme for charging for EARN sites then there is an incentive to reduce the umber of sites and the greater use of gateways. A side effect of gateways is to decrease the EARN traffic and increase the traffic which does not go through gateways - that is SMTP and X.400.
In principle decisions should be taken on technical merit. Funding arrangements should not affect technical decisions unless there is some generally agreed direction. Unfortunately this is not easy to achieve.
EARN is moving into a more competitive environment as there are an increasing number of alternatives - in particular TCP/IP mail and FTP are reaching an increasing number of alternatives - in particular TCP/IP mail and FTP are reaching an increasing number of places.
Members: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
1 Notes of meeting held July 31, TFOM/6/92
2 Progress on mounting PP
3 Progress on purchase of a Sparc station for PP development
4 Progress on moving EARN gateway to PP
5 Progress on use of NJE under UNIX
6 Progress on study of user interfaces
SMAIL study Studies in office automation group
7 Progress on study of alternatives to LISTSERV
8 Progress on draft strategy. TFOM/4/92
9 Progress on study of X.400 to EARN mail
10 Progress on study of address styles within PP
11 Progress with White Mail
12 Review of study areas
13 Date of next meeting
14 AOB
At Rutherford Laboratory there is a move towards the use of UNIX for scientific computing. As a result it is expected that the IBM computer will eventually be restricted to administrative tasks or phased out. Certainly it will not be available for EARN.
Rutherford is actively investigating alternative means of meeting those communications requirements which are currently provided by the IBM. These include the EARN gateway, LISTSERV and NETSERV. It is likely that an attempt will be made to move these services to UNIX platforms.
No decisions have yet been made and at this stage the Executive is asked to note Rutherford's intentions and is invited to comment. It may well be that other installations are engaged on similar activities and some coordination may be appropriate.
Rutherford currently operates an IBM 3090 with 6 vectors. The service is provided to a broad range of users including the scientific community and administrators. The PROFS office automation service and mail facilities are an important service.
The cost of running an IBM service is high and gives little flexibility in choosing computing power at an economic cost to match the user requirements. On the other hand UNIX offers a large range of opportunities from CRAY super computers to PCs with a measure of compatibility across the range. Much of the scientific computing is based on FORTRAN and is relatively easy to mount on a range of platforms. Other systems such as office automation and networking is rather more difficult to move.
A principle use of the IBM is as a large data store with large amounts of disk space backed by an automatic tape library. Means of retaining such a data store as a stand alone entity with FDDI connections is an important requirement.
Initially it was expected that scientific and communications uses of the IBM would cease within five years but this is now more like three years.
I have now been instructed to investigate moving such services from the IBM as soon as possible whilst maintaining the level of service.
There are many options but the key product is PP. PP is a UNIX mail system which will gateway between X.400 and Grey book mail. It is being installed in all UK universities to aid the transition to X.400. By happy chance the system will also gateway to SMTP. Rutherford is in the process of mounting the system for trials. Its use as a JANET/EARN gateway is an important aspect of this work.
There are two NJE systems for UNIX - UREP and HUIJE. These are being examined with a view to using one or other in conjunction with PP to provide BSMTP. The work is at a preliminary stage but looks promising.
There are a large number of mail list processors. Six have been identified and there are probably more. There is a LISTSERV 'look alike' which may well be a satisfactory replacement for LISTEARN. It seems to allow peered lists and several other aspects of LISTEARN. It is too early to know whether this is a satisfactory alternative to LISTEARN.
It is unclear how the lack of NETSERV could be overcome. The UNIX LISTSERV may be a possible approach or alternatively 'rent' a NETSERV on another EARB core machine.
There are several other aspects of the work which are not of interest to EARN such as the provision of mail stores and user interfaces.
It is possible that other countries or sites have similar problems of wishing to move to alternative mail products whilst retaining the valuable EARN facilities. If this is the case then some cooperation to ensure that good products are provided and maintained is important.
In the case of Rutherford there is the additional complication of being an EARN CORE site. We would certainly wish to retain that status and would also what to follow as closely as possible the spirit of the regionalisation project. Clearly some aspects would have to change - for example we would not be able to run VM, RSCS, FAL, and VMNET but would run UNIX alternatives.
At this stage the Executive is asked to note Rutherford's intention, to give such guidance as they see fit and to decide whether further activities should be initiated.
Members: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
Noted.
The reincarnation of "DIRECTORY" was progressing to schedule and PP would be mounted in the next few days.
The proposal (see TFOM/9/91) for a Sparc II had been approved by TMPC. R Westlake will get a formal quote and fill out a requisition. Action R Westlake
Waiting for PP.
HUJI has been obtained and could be mounted on the RS6000. Details of UREP have been asked for and are expected soon.
P Cahill will mount a number of systems.
K O'Neill is having problems with mounting CC. CC for WINDOWS has been ordered.
ISODE has no network layer and uses other products. Thus it can run over RFC1006.
There are 2 LISTSERVs for UNIX. A free one is available from Lyon and another from CS.B.EDU. J Wheeler is obtaining them and will mount. Action J Wheeler
No progress.
X.400 is available on the IBM and consists of three parts. OSI/CS which is part of VTAM, ONDS X.400 feature which is part of OSI/CS and X.400 PROFS. The cost is a monthly rental of £3K or a purchase of £125K (£30K after discount). A January 1 date for mounting this is not feasible. The implementation of RFC1328 for the gateway will require an unknown but large amount of effort.
Waiting for PP.
This has now been obtained and mounted. It waits for the provision of a message store which is promised as part of PP in beta test next month.
Nothing to report.
September 17 - 9am.
Nothing to report.
£6478.95 is outstanding on this order due to NSCs failure to supply the equipment. They have supplied alternative equipment manufactured by DEC which is a satisfactory substitute. Thus I am happy that the invoice is now paid. We had expected NSC to substitute their produce but this has not yet happened. We are now wanting an upgrade. We are negotiating with NSC on the basis that if they cannot replace the equipment by the end of September then we will keep the NSC equipment and subsequently upgrade it via an order on DEC. I am happy with this situation and we should now pay the bill.
Dear Vikki,
Thank you for your quote for two extra FDDI concentrator ports of August 14.
Further to our telephone call may I ask you to confirm that since you will be unable to deliver NSC equipment by the end of September you are happy for to retain the DEC concentrator permanently on payment for two further concentrator ports?
To progress quickly I propose to issue an order based on your quote and will expect you to invoice us at your early convenience. Please note on the invoice that the equipment has been delivered to avoid payment delays.
Thank you for your help.
Yours sincerely,
Paul Bryant.
To: Miss S A Gill From: P Bryant Ext: 5267 F8, R30 Date: 15 September 1992
Just back off my holidays. Thank you for your memo. Here is the information you need:
Alister Simpson Cabletron Systems Ltd. Network house Newbury Business Park London road Newbury RG13 2PZ (0635) 580000 Vikki Ames Network Systems Limited Kings Ride Court Ascot SL5 7JR (0344) 23399 Mick Mazuryk Datarange Communications Limited Kingsmead House Abbey Barn Road High Wycombe HP11 1XG ()494) 441256 Jane Johnson Lanbase plc Solutions House Berkshire Business Centre Berkshire Drive Thatcham RG13 4EW LattisNet SynOptics Communications House Station Road Cookham Rise Berkshire SL6 9BU (0628) 819330 John Donovan 3Com (UK) Limited Pacific House Third Avenue Globe Park Marlow SL7 1YL (0628) 89067 Eric Case BUS-TECH The Wokingham Business Centre Ltd The courthouse Erftststadt Court Denmark Street Wokingham RG11 2AY (0734) 894800 Steve Ash Cisco Systems Ltd Unit Three Cliveden Office Village Lancaster Road High Wycombe HP12 3YZ (0494) 464944 Bill Atkinson Computer International Ltd. 10 Farmbrough Close Stocklake Aylesbury HP20 1DQ (0296) 436965 Peter Chambers Digital Equipment Co. ltd. The Crescent Jays Close Basingstoke RG22 4BS (0256) 370000 Mike Todd European Information Technology Bennet Court Bennet Road Reading RG2 0QX (0734) 314104
B) Attached
C) The Req on DBS is going through the system. I attached a copy that still needs its project number and raft of signatures
D) Software. Difficult. These "things" come as systems which are part hardware and part software and the two are intimately connected. Thus the cost will include software and we will expect maintenance of the software with the maintenance agreement for the system. I hope this answers the question.
E) The existing network to which the equipment will be directly connected is our own and as far as I know all such equipment is owned and not rented. The equipment will be indirectly connected to equipment owned and rented by many parties as this network will be connected to JANET.
Paul.
Members: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
Apologies P Cahill and K O'Neill
Noted
This has been mounted and tests are under way.
Now on order.
Without NJE on the PP machine it is essential to retain a MAILER on the IBM. It may be possible to use TCPMAIL on the IBM instead. These could be unmodified versions.
Can TCPMAIL be used? Action G Christian
There may be merit in moving to international domain ordering as a number of UK universities have. Can this be done and is it sensible. Action J Wheeler Action D Parker
HUJI is multistreaming.
UREP is still impossible to obtain. EARN may try and support UREP. Action P Bryant
HUJI is now on the RS6000. Tests will take place. Action T Kidd
The UNIX Andrews system which is "all singing and all dancing multi media" is being looked at. It is very large. Action K Hoadley
No progress with CC. Action K O'Neill
NUPOP on the PC looks very promising. Its main problem is that it needs a user account on the PP machine (or another convenient one). If this is recommended then it may be necessary to have a number of village machine providing the message store. Action P Bryant
The UNIX versions will be put up and evaluated. Action K Hoadley Action J Wheeler
It appears that Eric Thomas has a low opinion of these products and wants to write a new one. EARN is looking into this. Action P Bryant
Apparently we can have access to MAILBASE. Unfortunately it needs INGRES. We need to decide whether to get INGRES to test MAILBASE. This depends on costs. Action R Westlake
No progress.
Note - OSI/CS is an option under VTAM.
No progress is envisaged. It is still considered that an implementation of RFC1148 on the IBM is a very large task.
Progress will be after a workshop on Monday and Tuesday of next week, K Hoadley, G Christian, P Girard and R Westlake will attend.
Awaits message store next week.
Nothing to report.
October 5
Nothing to report.
Particle Physics Jason Leake, R1
Informatics Andrew McDermott, R1
Space Science Chris Mutlow, R25
Science Richard Brodie, R3
Computing Brian Day, R31
Roland Brandwood, R31
Kevin Hoadley, R30
Chairman Paul Bryant, R30
Secretary Jeremy Sharp, R28
2 Delivery time scales
3 Equipment location
4 Cables
5 Local FDDI networks and end equipment
6 IP addressing
7 Date of next meeting
8 AOB
This paper is some initial notes on how a conference organisation system may work. It is not complete. Is this a sensible project?
At the centre of the system there is a database containing all the data for the conference.
Access must be by a variety of means. Full screen access via X-windows would be best for browsing and modifying the database although line at a time is needed for those with more primitive terminals. Mail access is needed for group communication. FTP or LISTSERV mechanisms are needed for access to mail and abstract archives.
The data base contains:
People - name - address - mail - address
Delegates - pointer to "people" - pointer "hotel" - paid ?
Organising committee - pointer to "People"
Programme committee - pointer to "people"
Other committees - pointer to "people"
Speakers - pointer to "people" - title
Session - pointer to "speakers" - title
Chairman - pointer to "people" - pointer to "session"
Agenda - pointer to "session" - pointer to "room" - time - duration
Room - name - size
Hotel - name - address - telephone number - number of rooms - booked
The programme will ask questions to set up the data base as far as possible and will also set up a draft time table.
For example, after the projected data is input the program will assume the organisation will take a year and will define dates by which activities should be complete. When the duration is defined a draft agenda will be set up with 4 sessions per day and evening activities.
Any assumptions must be capable of overwriting, for example, to set up fringe meetings, demonstrations and social activities.
The conference set up must be easily displayed and changed.
Various reports must be available from pert charts to lists of delegates.
It must be possible to ignore parts of the program, such as hotel details, so that these can be undertaken by other organisations.
Distribution lists will be set up for all the groups and for the totality of groups. An archive will be kept with an index.
There will be an archive for abstracts and other text for calls for papers and application forms.
The expenditure after six months is within budget and shows an expected underspend of 4.7%. Note that this takes account of the reduction of the salary paid by EARN-France from 46 to 23.91 KECU. The out-turn estimates are pessimistic and actual expenditure from more recent figures show an underspend of about 9.5%.
The principle area of underspend is within the development budget where plans have not matured as fast as expected. In this area a number of projects have not progressed to the extent expected and it now seems that no EARN Scholarships may not be awarded. Members may examine the figures in more detail when the relevant paper is discusses.
The 1993 budget is for approval. It has been amended according to the decisions of the Innsbruck Board of Directors meeting. In the light of EARN's lower budget level and the healthy reserves no contribution to the contingency fund is proposed. It should be noted that the budget has been drawn up on the assumption that EARN will continue during 1993 as it has in 1992. There are a number of uncertainties in this assumption. In particular it now seems that the Operational Unit will not provide any NJE services in 1993. The support of RIPE and EBONE are in line with previous decisions of the Board. Members will have heard the President's report and may wish to examine any financial implications in the budget discussion.
As decided at the last Board meeting, countries may have a subscription based on GNP (as now) or on the number of sites within the country. The subscriptions in the budget paper assume that each country will pay the minimum of the GNP and site subscriptions.
EARN has for many years attempted to determine a "fair" subscription scheme. Members will remember that some years ago it was expected that subscriptions would eventually be based a a fixed part and a part dependent on traffic levels. This was rejected by the Board after technical problems with collecting figures and problems accounting for server traffic. At that time there was some suggestion that EARN would eventually pay for all the lines. Things have changed and with the use of IP connections such ideas are out dated. The introduction of a site subscription is perhaps a new direction where EARN may eventually follow CREN and move to being based on site memberships. This may well better fit EARN's concentration on support services, higher level services and an annual user directed conference rather than on the provision of NJE and lines. Indeed, it is clear that many sites profit from EARN which have no direct EARN connections. Such moves need to be considered in conjunction with how EARN might align itself in the future. There have been suggestions of mergers or separation of tasks with several organisations over the years (EUNET, RARE, the Operational Unit and CREN to mention a few) which would affect subscription models. The Board may like to consider the subscription models that the Executive should study for 1994.
Unfortunately a number of countries are still not paying their subscriptions on time. The Executive have suggested various encouragements for early payment, for example - discounts for early payments or premiums for late payment, which the Board has not adopted. Thus we can only urge countries to try and speed payments.
Distribute: J Barlow, P Bryant, T Daniels, J Hutton
AGENDA
1 Minutes of meeting held 21 July, 1992.
2 Matters arising
3 Recent performance
4 Diversion of mail from BSMTP to SMTP
5 EARN regionalisation project
6 Connection to Germany
7 GBOX
8 Finance
9 Board of Directors meeting
10 Date of next meeting.
11 A.O.B.
There has been some progress. Progress has been less than expected as there is currently only a modest user demand. Secondary reasons are the lack of equipment and lack of software - both need some funds.
The project proposal (IPTAG/92/6) is still the definitive description of the project.
The NSAP proposal (IPTAG/92/37) is a provisional definition of the NSAPs to be used.
Considerable progress has been made in the DECNET community with phase V testing. (DECNET phase V uses CLNS at the lower layers and also uses several ISO application layer protocols). The work has been centred at CERN and Rutherford Laboratory has cooperated in the enterprise. The routers used have been the DECnis 600. Routers are now being loaned to Cambridge, Manchester and Oxford to undertake more extensive testing. It is likely that a case will be made for a set of DEC routers for the HEP community. sites.
The DECNET work is complicated by the need to provide DECNET phase IV to phase V conversion facilities which are embedded in the routers. It is unclear as to whether other router suppliers will provide this facility.
DECNET further complicates or restricts options as all the machines need to be in the same routing domain until phase IV vanishes.
Although the DEC community would like to see a CLNS network for general rather than exclusively DECNET use it is so far unclear the restrictions that this may place on such a network. For example, it may demand the use of DEC routers.
The principle non DEC ES product is from SUN. Unfortunately this requires money and has not yet been purchased at Rutherford.
In cooperation with Edinburgh some testing has been carried out with a PC ES system which currently extends to transport layer. This has been a very useful educational exercise.
A Cisco router became spare at Rutherford a few months ago for the project. Unfortunately it cannot run the latest Cisco software and the current software has a number of bugs. None the less it is being connected to Edinburgh and we will test the PC software initially. We also hope to try DEC and SUN products.
To progress further some funding is needed.
The participating sites need the SUN CLNS software. This includes FTAM but it can also support ISODE higher layers. The cost is £1000 per machine but in the light of the experimental nature of the project it may be possible to reduce this cost. Two licences per site would be useful to allow testing between machines on a local area network as well as across JANET.
Rutherford need to upgrade the router being used as this cannot run the latest CISCO software. It would be of limited use to spend time re-discovering bugs fixed in later software releases. This will cost approximately £3.5K plus VAT. It is believed that other sites do not need router upgrades.
Interworking testing with DECnis600 to Cisco routers will take place. When this is complete it will provide a path via CERN to the European CLNS project.
If and when a connection is made via ULCC to Europe some care will be needed to ensure that the DECNET phase V traffic is not disrupted by, for example, sucking all the current phase V traffic through ULCC.
We hope to connect to the other interested sites at an early date.
The funding required for equipment for sites is less than £10K and could be less if SUN are wiling to reduce there costs for software.
Now that the local area network plans are well launched attention has turned to mail.
Mail is a very complicated area. In the past it seemed simple with a nice well defined path to use X.400. Unfortunately it is not so simple. We have two major mail protocols - X.400 and SMTP - both of which look like surviving for the foreseeable future. We also have Grey book, which although expected to be phased out, shows little sign of a quick departure. The situation is made more complicated with the many workstations now installed and by products such as Novell Netware which have a measure of popularity.
To this end we have set up a working party (The Future Of Mail) to study the problems and propose a strategy. So far the problems are fairly well identified and quite a bit of practical work is under way to evaluate possibilities. In particular PP, the answer to a networker's prayer, supported by the JNT is a strong possibility to play an important role.
2.1.1 Coloured Book Services (JTMP, NIFTP, SSMP)
Stable apart from occasional JTMP problems.
2.1.2 Network/3270 and line-mode terminal services
Stable.
2.1.3 Interlink DEC/IBM gateway
Version 1.3 of the Interlink package (with pre-release 1.4 DECPVM and DECGAS) was installed with few problems. The long-standing DECPVM storage overflow bug has gone.
R1 users get intermittent delay problems when using DECPVM in some circumstances, but this is not a consequence of the new release. Attempts to reproduce the problem outside R1 have always failed, leading to a view that local configuration parameters in R1 DEC equipment may be to blame. Still being investigated.
2.1.4 VTAM
V3R3 of VTAM continues to be generally reliable.
2.1.5 IBM 3745
The software has been upgraded to NCP V5R4, EP R9 and NPSI V3R4 with few problems. This is a great improvement. The previous release took several weeks to debug, and the release before that took many months.
The link to the Netcomm has been upgraded from 64K to 128K, and the link to the JPSE has been removed. There is still a 64K link to the SEEL.
2.1.6 Configuration
The conversion of the 3174 Clusters to SNA is complete and the CLUSTER virtual machine (running PVM) has been scrapped.
EP is still present in the 3745 at present to support the VM Telex facility.
2.1.7 X.25 and Pink Book
VMNCP 7.5 is still the latest released version, driving X.25 (JANET) through VTAM and NPSI, and Pink Book (Ethernet) via the IBM 8232. Production trials of the preferable 'Fast Connect' interface between VMNCP and NPSI have taken place but have as yet failed to survive for more than about 8 hours in prime shift.
2.1.8 RSCS
We continue to run Version 2.3.05.
2.1.9 TCP/IP (including the IBM 8232 and the NSC 7000)
Version 2.2 is now being configured and will become the production service in a few months time.
The NSC FDDI kit has worked without problems. A DEC FDDI concentrator has been installed into the ring ready for the RS6000 and the new VAX kit.
The VMNET (RSCS NJE over TCP/IP) service now supports ten links and continues to operate without problems.
2.1.10 Other IP Based Projects
The Virtual Tape Protocol is being used by an increasing number of people. Recently EISCAT put copies of their 2000 tapes into the silo and are accessing the data using an application program written on top of the VTP library.
The VTP library has been ported onto VAX VMS, the "tape" command is being converted to use VMS style options and handle VMS file types.
The work on an experimental network calendar server which provides access to the OV/VM calendar engine over an IP socket-based interface has "stalled" due to pressure of other work.
The IP-based SYSREQ client has been implemented on the Cray X-MP, a number of Unix workstations, VAX VMS and PCs.
A batch facility for TCP/IP FTP on the IBM 3090 is being developed by Jeremy Sharp (from Telecoms Group).
Some time has been spent configuring IP on the Cray X-MP and Y-MP machines including putting the Operator Workstation and Maintenance Workstation onto the site ethernet for a short period (whilst the Y-MP was being configured).
2.1.11 MAILER
The Mailroom facility is ready for use.
Code has been written to reject mail containing origin addresses that parse incorrectly and hence cannot be reversed.
The reversal of foreign addresses by the EARN side of the gateway has been simplified.
More remote partial domains that were handled by EARN are now being reached via IP.
Investigations have started into the possibility of replacing the MAILER and EARNGATE virtual machines by PP running on a SUN.
2.1.12 RS/6000
Responsibility for network support on the RS/6000 was formally taken over at the end of the reporting period, coinciding with its introduction into service. This includes TCP/IP, X25 and JNT Coloured Books.
2.2.1 Migration to the VM/ESA version of GCS is likely. This is a prerequisite for installing future releases of VTAM and for running GCS applications in XA-mode. A new release of SSP should also be installed.
2.2.2 The 'Fast-Connect' method of using NPSI should be either finalised or abandoned. It seems to be exceedingly difficult to remove the remaining subtle problems.
2.2.3 A new release of the Coloured Book software for the RS/6000 should be installed. This is a prerequisite for moving to AIX 3.2.
2.2.4 Work on RSCS Version 3 should be completed and the new system installed towards the end of November.
2.2.5 TCP/IP version 2.2 will be installed as the production system in October.
2.2.6 The updating of NFS mapping tables on the Cray should be made automatic.
2.2.7 Mail that comes in over JANET should be checked for a valid NRS address in the From or Sender header field.
2.2.8 It will be possible to send mail from PP to EARN without using EARNGATE and MAILER.
3.1.1 Local
* There has been no progress with the replacement floor in the Telecomms area.
* The Operations section has been short staffed during the period due to annual leave and lack of recruiting, hopefully this will be resolved in the next quarter. However, when necessary, staff from other sections have been called upon to give a helping hand and this was required for some days following the thunderstorms early in August. This has meant that other jobs have been delayed.
* The storms over the weekend of 8/9 August caused a considerable number of problems and outages.
a) Main problem areas were R25 and R1, however there were problems all over the site.
b) In R25 a number of synchronous linedrivers required replacing, a number of PAD boards were replaced following damage to RX/TX and Baud generator chips. A Pilkington multiplexer had to be swapped out. The ethernet in TechElec was also badly disrupted and assistance was given to help them overcome their problems, also equipment was loaned to them. A VAX had its DMV board damaged and the SEEL interface to which it was connected also required replacing.
c) In R1 the power to the Telecomms racks was not restored until the Monday morning. This meant that users associated their problems with this and when the power was restored it took the users sometime to report problems. A Focom multiplexer had 6 out of its 8 channel boards replaced. PAD problems were restricted to individual connections and these were swapped to spare ports and the problem dealt with later.
d) Atlas Centre. Two asynchronous linedriver boards had to be replaced as chips were burnt out. A board in the SEEL was also exchanged.
e) There were general PAD, Multiplexer and linedriver problems in other buildings.
f) Reports of problems associated with the storm were still being received after a couple of weeks because users were either only infrequent users or had been away on holiday. In numerous cases the reported problem had two faults, the terminal and the piece of equipment it was connected to.
g) There were a high number of terminal faults reported which had to be fully investigated.
h) Most of the old equipment damaged will not be repaired or replaced. To do so would be uneconomical, the current holdings and use do not warrant the expenditure.
* The PSEs have been reliable during the period.
* The Ethernet Bridge for the Services Village had to be changed following problems associated with the air-conditioning shutdown. The replacement Bridge unfortunately was not up to the same mod level and further problems were experienced with broadcast storms.
* The Payroll equipment was trouble free but changes made at Chessington caused interruptions to the service.
* The CERN line had two outages, one for 12 hours. Both were due to problems within France. The X.25 link to CERN also had numerous problems but these were due to problems with CERN equipment.
* The line to Brussels was plagued for a period early in the quarter, again most of the problems being the responsibility of the remote PTT.
3.1.2 JANET
* The GPT 4190 (JPSE) was closed down on Tuesday 11 August, all the connections were transferred to a new Netcomm 3500 which is called Atlas-1.
* The current JANET switch configuration at RAL is as follows.
a) Atlas-1 Netcomm 3500 9.6K connections and some 64K. Connects to Atlas-2 for the rest of JANET.
b) Atlas-2 Netcomm 3500 2M and 64K connections. Main switch for UFC supported connections, JIPS, other JANET switches, RAL SEEL, RL.VMSFE and RL.IB.
c) Atlas-4 Netcomm 2000 supports PSS Gateways, JANET.NEWS and RL.GEC-B. Connects to Atlas-2 for the rest of JANET.
d) Atlas-5 Netcomm 2000, ECB for JNT ethernet plus other JNT systems. Connects to Atlas-2 for the rest of JANET.
e) The 3500 supports 32 connections and the 2000 12.
* Atlas-2 suffered from overheating during the air-conditioning shutdown. The venting systems have been modified to ensure a good airflow.
* The following sites have had their line speeds increased to 64K, Plymouth University, De Montfort University (Leicester Poly), Oxford Poly. and ADF Farnborough.
* Lines to Rennel Centre and Oxford Atmospherics have been cancelled. Both sites now come through the local university.
* The main BT faults have been the re-occurrence of problems on the second Kilostream line to Wallingford, and with the line to Wormley going out of spec.
* The Mercury line to OCLC Birmingham gave high error rates, the fault being the cable from the Mercury equipment in the basement to the 7th floor of the building. Both Mercury and the JNT said the other was responsible for the cable. In the end the line was cancelled and a BT line installed. There have also been problems with the Mercury line to Swindon.
* A PC with X.Windows on it has been installed in the IBM operators area to allow them to get better view of the JANET network.
There were 97 reported terminal faults in this period. The sharp increase is due entirely to the electrical storms of the weekend of 8th August. These damaged the interface chips of many of the terminals in R25, R1 and R2, along with the pads and multiplexers associated with them. This caused a backlog of faulty units while spares stocks were replenished.
The sub-section staff repaired 23% of the faults, the remainder were done by maintenance agency.
The distribution of agency repairs is as follows:
Last qtr. This qtr.
KODE 54 75 Contracted terminals
OTHER 12 0 Warranty repairs
The period's faults have been analysed for divisional distribution. Of this total, 21 (20%) were CCD and the remaining 80% for other RAL divisions.
_______________________________________
| | REPAIR AGENCY | |
|Division |____________________| |
|-Project | TCOM | KODE | OTHER| TOTALS |
|_________|______|______|______|________|
| ADM | 4 | 2 | | 6 |
| BNSC | | 1 | | 1 |
| CCD | 1 | 20 | | 21 |
| CO | | 1 | | 1 |
| EBW | | 1 | | 1 |
| INF | | 5 | | 5 |
| PPD | 1 | 1 | | 2 |
| RCR | | 1 | | 1 |
| SCI | 4 | 19 | | 23 |
| SL | | 1 | | 1 |
| SPA | 8 | 9 | | 17 |
| TEC | 4 | 14 | | 18 |
|_________|______|______|______|________|
| TOTALS | 22 | 75 | | 97 |
|_________|______|______|______|________|
Work on the SQL databases is on-going.
Kode International has bought DCM Services Ltd. and with it has merged its maintenance operation, Kode Computers. The new agency will trade under the banner of DCM. There will be some changes to the operating practices, but it is hoped there will be no degradation of services during the coming change-over period.
A thin ether segment was extended in R31. The topology of a thin ether in R68 was changed.
A fibre optic cable (FDDI) was installed in Atlas for an RS6000.
Several coax cables were installed between R1 and R25. More coax cables were installed in R2.
An AUI cable was installed for the relocated CrayFE. An AUI cable was installed for the XMPFE.
Offices in R12 were cabled for the relocated Safety Section.
Cables in R20 were identified ready for them to be connected to R68 in order to provide comms for the remnants of R20 that will exist after October.
Appletalk cable was installed in R1.
As usual coax, PAD and DECserver connections were shuffled around.
Hundreds of redundant cables were removed from Telecomms.
Cables in R20 will be routed to R68. Cabling will be installed in R12 for PAD, ether, Appletalk & '3270' links, Coax cables between R2 & R3 will be re-terminated to overcome wiring faults.
The Services village will be extended into the West computer room.
A fibre optic (FDDI) cable will be run from the Lower computer room to the west computer room for a VAX.
Thin ether will be extended and modified in R25 (3), R2 & R65 and may be modified in R3.
Two thick ether cables will be installed between R25 & R68. A thick ether cable will be installed between R65 and R12.
Appletalk cabling will be installed in R25 and may be modified in R3.
In early June, PPD and CCD has successfully jointly organised a UK-wide HEP/SPAN DECnet meeting at RAL. The meeting brought an update to the audience the recent development on DECnet/OSI in general, the HEP/SPAN pilot project in Europe and the activities at RAL in particular. At the end of the meeting, it was agreed to launch a UK based DECnet/OSI pilot project for the HEP/SPAN community, in the hope of bringing the transition as smooth as possible.
6.1.1 Network Software
A new WAN Device Driver WANDD V1.2 has been released. This provides the support for the synchronous devices for newer types of machine (VAXStation 4000 and MicroVAX 3100).
There had been no software updates on PSI, LLC2, CBS and JTMP, but a further fix kit to DECnet VAX Extensions was received in early June. This update fixes a few major problems including one that had caused system to crash when communicating between a Phase 4 and a Phase 5 host over X.25. In late August, a pre-release of DECnet/OSI for VMS (Wave 2) kit was installed on the CCD Development VAX. A number of problems were found and reported back to Digital. It is hoped that these problems can be resolved before it is released in October.
The monitoring of active DECnet hosts and connections and the registration of the machines in the RAL DECnet area (Area 19) continued. Currently there are 838 active machines out of a limit of 1023 allowed and 442 of them have provided us their networking details. The information collected will be useful for planning the DECnet/OSI migration for the UK HEP/SPAN community.
From 13 July, the GBGBOX link to ITGBOX (Italy) link has terminated. All Italian traffic are now being routed through the VMNET on the IBM.
6.1.2 Network Hardware
The IP address for the IBM has been reconfigured on all the CCD VAX systems to reflect the use of the FDDI network.
A new Cray station link was set up on the VAX front end (CRAYFE) to talk to the Cray YMP8I/8128 in late August. The new station is currently running in parallel with the existing the link to the Cray XMP.
A field test has started on a Digital Alpha Cobra VAX (RLVH) in late June. The field test is aimed to help migration of the VMS applications to the Alpha Open VMS platform. DECnet and LAT have been installed on this machine.
No further changes in network hardware on the CCD VAX systems were made in the past quarter.
The regular networking support of PSI with PSI Access , CBS and Red Book, Pink Book, DECNet, LAVC, PSS/IPSS, VMS-COMMS, EARN and Janet mailshr continues. The NRS updates have been continually supplied to SERC sites. Work continues to short list firms who could supply maintenance of Digital hardware and software on site.
No major security problems have been recorded in this quarter.
Digital is still in the process of negotiating with UWCC to renew the CBS contract, which has been described by Digital to be "close" to completion. When that is in place, the VAX group will try to participate in the field test of the next CBS release which will provide the VAX clustering support.
A VAX 6000/620 machine will be installed in CCD in mid-September. This will be offered as a service machine and in due course will be upgraded to an Alpha based machine. It is planned to install DECnet, LAT, PSI Access over Ethernet and TCP/IP over FDDI as soon as the machine is delivered.
The field test on the NIS 600 router and DECnet/OSI for VMS (Wave 2) will continue. Provided permission is sought from the JNT, the work, which also forms part of the UK HEP DECnet/OSI pilot project, will be extended to across the site over JANET, to Cambridge, Oxford and Manchester.
Talks will continue with Digital and others for a tender to provide software and hardware maintenance on DEC equipments for next three years.
The DECnet nodes monitoring and registration will continue.
7.1.1 NJE
There remains a single connection via the GBOX which goes to Germany. VMNET links now exist to all of the other EARN core sites and this now includes CERN.
The completion of the core network and the introduction of EBONE has made a significant improvement to the EARN performance. Traffic now normally only takes minutes to be delivered and large queues are rare. EARN now produces much better statistics which measure queue lengths on core sites and also end to end file transit times.
Traffic levels are stable. The removal of mail traffic with good domain addresses made far less impact than expected.
In total the EARN traffic is still increasing although there is little or no increase in UK traffic. However, since non EARN IP network traffic is showing a significant increase the lack of increase with EARN is some indication that EARN is possibly towards the end of its life.
7.1.2 File Transfer Between EARN and JANET
No change.
7.1.3 Mail
See section 2.
7.1.4 LISTSERV
There have been a number of problems, particularly with updating LISTSERV, which have been irritating.
7.1.5 EARN Management
EARN has been considering its relationship with RARE, RIPE, the OU and EBONE. There is some indication that EARN may well start to concentrate on end user services and in particular documentation. No firm recommendations are yet in place. EARN now produces an electronic newsletter and will help to produce a joint RARE/EARN newsletter - TELETRIBUNE.
Hopefully the funding mechanisms for 1993 are now sorted our. It is clear that further changes will be needed in 1994 and thoughts are directed to following CREN and going to a site based charging scheme for all subscriptions. This would help a transition to concentrating on end user services.
7.2.1 NJE
VMNET connection to Germany.
RSCS version 3.
7.2.2 File Transfer Between EARN and JANET
No change expected.
7.2.3 Mail
No change expected
7.2.4 LISTSERV
A major upgrade is expected.
7.2.5 EARN Management
The BOD meeting will be held in November to agree the budget and subscriptions. This will take place at EARN's conference - Network Users Conference - in Pisa.
There are thoughts that the next Network Users Conference could be in the UK.
The statistics for PC orders over the last three months were 148 orders placed, which included 44 PCs, at a total value of £169,859 including VAT. The number of PCs ordered is the same as last quarter despite the summer holidays and the grand total is now 747, up 184 on this time last year which is another record for the annual increase.
Thus for yet another quarter PC Support have been rushed off their feet installing machines and upgrades at what is normally a quiet time of the year. It is to be hoped that last autumn's jump in orders is not repeated this year otherwise we might not be able to cope. The repair workload has also not slowed during the summer months with a steady stream of broken screens and system units. This has not been helped by the continued unreliability of IBM PS/2 machines with several suffering from a failed hard disc or motherboard. The lightening strike also damaged a lot of PCs mostly on the asynchronous communications card which is fortunately very cheap ( £10 except for PS/2s which are £60) and we had plenty of spares. A Compaq PC was also severely damaged.
The CVN for a new member of PC Support had a closing date of 21 August but as yet nothing has been heard from Personnel as to whether suitable candidates have been forthcoming.
The process of persuading users to contact Service Line with their PC problems continues to be an uphill struggle. We now have access to the TFAULTS database and a lot of effort has been put into clearing off old IRs. Unfortunately a new database is being phased in so the system is not as helpful as it should be in the future. A document is also being produced to guide Service Line in the difficult task of deciding to where a PC fault should be directed. The general break down is that applications go to OSG, communications to either Telecoms or the local Ethernet Village Manager and PC hardware and systems problems plus everything else to PC Support.
PC prices continue to tumble and it is proving difficult to keep up to date. By the time we receive the printed version of one month's prices the end of the month is approaching and a phone call to Viglen reveals next months prices are even lower. Viglens continue to be very good value for money being among the cheaper good quality suppliers though, with the continuing game of leapfrog between the manufacturers, they may not be the cheapest at any given time. The specification is also improving with all Viglens now coming with improved video chips giving a 235% video speed increase at 256 colours and 1024*768 resolution according to the video benchmarks. They also support 1280*1024 resolution with 16 colours. Viglen now have models incorporating the new Intel speed doubling chip whereby the internal chip speed is double that of the rest of the PC. A 50Mhz version has been tested and, although it gave a significant CPU performance increase compared with the 25Mhz 486, comparisons with the 33Mhz variant suggest it was not worth the price premium for the overall gain. The 66Mhz version has just been announced which is now the fastest machine they produce. It is unfortunate that Viglen have not brought out a standard 50Mhz machine which is likely to give a better all round performance.
The rapid rise of the size of packages requires a regular review of the standard disc size purchased. With around 15Mbyte being the normal requirement for a Windows based package a 40Mbyte disc is far too small and 100Mbyte is the current standard. The small additional cost (£95) for a 200Mbyte drive is also easy to justify and this is rapidly becoming the norm. Another bottleneck on the PC, especially noticeable under Windows, is the video bandwidth and some of the cheaper solutions are being evaluated. A TIGA card was purchased some while back but this has now been eclipsed by the improvements of the current Viglens. However the appearance of cheap S3 bus cards, which are specially tailored to the Windows environment, mean that further improvements may be possible for a reasonable outlay. One has therefore been purchased for evaluation. Work is also planned on evaluating the multimedia facilities in Windows 3.1 and sound cards and a cheap CD-ROM player have been ordered .
PC Support now has use of the OSG Novell Netware server and this is giving valuable experience in using such a network. However the small size of its disc and the limit of 10 users mean it is not suitable for file sharing within the section. A 1.2Gbyte SCSI disc has therefore been purchased and will be initially attached to a Sun 3/50 for use as an NFS server. The likely development of a PC version of Unix which includes Novell Netware support has been noted and should this materialise at a reasonable price then the disc could be moved to a large PC for it to be evaluated.
There was only one virus incident in the last quarter and it was local to one machine so was not serious. We are now paying for a more comprehensive anti-virus service in order to try and reduce the staff costs of keeping up to date in this area which currently runs at 0.1MY.
Version 1.2 of PKTMUX, the program that allows multiple IP protocol stacks to run over one ethernet card, is now in beta test. As well as fixing several bugs it now supports 802.3, thus allowing ISO protocols to run alongside IP. It is also much more efficient on certain ethernet cards because it can now locate the data buffer in the Packet Driver and avoid copying packets in a lot of cases. PKTMUX is in use around the world and this has made a significant contribution to sorting out problems with it.
A lot of effort has been put into making the TCP/IP variant of MOS v2.3 run under Windows 3. A problem with the screen only partially updating has still not been solved but now occurs less frequently. The development of the PCTCP version has consequently been delayed and when this is done version 2.3a will be released.
PKTMUX v1.2 should be released.
A driver to run MOS v2.3 over PCTCP should become available.
Recruitment of a new member of staff as quickly as possible.
The IPTAG project to evaluate CLNS is now making some progress with the RAL Cisco now connected to an ethernet spur. Testing of a PC implementation of CLNS has been done with some difficulty. K Hoadley as well as being secretary of IPTAG is also the secretary of an offshoot group SuperJIPS which is considering the problems of running IP over SuperJANET.
The section has continued to contribute toward the planning and tendering exercise for the purchase of a site FDDI service.
LPD development has continued and the product is now more robust. Documentation has been produced and an LPD service has now been installed in the Directorate and appears to be working well apart from hardware failures. The testing of LPD has highlighted glitches in the various components involved in printing from a PC and the perception of what is a good product has been marred by factors outside its control. For example the printer redirector in PCTCP just does not work at all on some days, and the various clones of Adobe's postscript interpreter will not print certain files. Postscript graphical output can also exceed the available memory on some printers. Fortunately a normal user is unlikely to hit most of these problems but it does highlight that this area is not without its difficulties.
Progress in the merging of MACREG into the DNS has been slow due to the pressure of other work. It is proposed to show the basic system to the LMC for their comments during the next quarter. Operational documentation on the DNS is nearing completion and should soon go to Operations for their comments.
A PC program (PDTBUILD) for producing tables of IP and MAC addresses has been in use during this quarter. It has proved useful in showing what is actually running on the network and can be used to locate duplicate IP addresses. However it has no automated means of signalling such occurrences so can only be of help after a problem is reported.
K Hoadley has spent a lot of time with R Westlake rebuilding the system on the Directory machine (nee Thomson) and mounting ISODE, QUIPU, PP, X.25 and Coloured Book software. It can now function as a mail gateway and some testers are evaluating its use. During part of this time the Directory machine was unable to act as the primary DNS and this caused some problems as the secondary DNS, Hardy, proved unreliable for reasons which are still not clear. This also highlighted problems with some of the IP software in the area of DNS addresses and this is now being reviewed. A mailing list of the users of each of the supported IP packages is to be set up so that they can be rapidly notified of problems such as these.
Approval has been given for the purchase of a network services machine and a Sun SPARCserver II has been ordered. It will be initially used to provide the PP service and act as a secondary DNS.
A steady demand for PC-NFS and Vista eXceed Server installations has taken a lot of effort. Eventually, like PCTCP, such installations will be done by PC Support once sufficient experience has been gained to write the necessary documentation. The form filled in by a user when he requests an ethernet connection is also being revised in the light of experience. The aim is to make sure that the user gets competent advice on what they really need and also has the approval of the Village Manager for whatever is being run on the PC. The problem of keeping networking software up to date has also been addressed. The frequent arrival of new versions means it is impracticable to install them on each user's machine so it is proposed to advise users of new versions via the mailing list so that they can request an upgrade if they need the new facilities and /or bug fixes provided. When the PC Support NFS server becomes operational an automated means of installing and upgrading networking software will be investigated.
A draft copy of the user documentation for PCTCP has been produced and one for PC-NFS is in preparation. Progress has not been helped by the changes required for new versions of the software but it is hoped to get copies out to users this quarter.
User documentation for PCTCP and PC-NFS should become available.
None.
To be supplied
VXRAL VAX has been successfully connected.
The RS6000 interface has been delivered and now awaits software from IBM which can support FDDI and X.25 concurrently.
The VME bus testing has not progressed as this has a relatively low priority.
As a result of NSCs inability to deliver their concentrator the DEC concentrator will not be replaced. The spare two channels are being purchased from NSC. A further four channels are being purchased.
The working party to produce a new local area network strategy has produced its report which has been accepted by TMPC. Five villages are now interested in taking part in an initial FDDI installation. £100K has been available for a FDDI installation and a tender exercise is now underway to select a supplier.
A further 4 concentrator ports will be delivered.
Further FDDI connection will be made from VAX equipment.
The RS6000 connection will be tested and brought into use.
The VME board will be tested.
The tender document will be issued to suppliers and responses will have been received.
An installation working party will have been set up to install the FDDI network across site.
The way in which the site FDDI will be connected to Super JANET should be clarified.
As the cost of FDDI equipment drops so several machines are expected to be connected.
The need for monitoring equipment is being studied but as yet there is no recommendation.
12.1.1. IBM
12.1.1.1 UNIXFE
Telnet hangs into UNIXFE persist and attempts to fix this have failed and merely pointed to the poor support available from our VAR. This culminated in an abortive attempt to install AIX 3.1.7 which required a full restore of the system after this revision killed X.25 access.
SNMP was configured and tested however it was decided that a simple ping from PROP on the IBM was sufficient to determine availability.
X.25 documentation completed.
UNIXFE lost one of the system disks in mid June and it transpired that the backup precautions were not adequate to make restoration easy. A standard inode backup, because of the peculiarities of the LVM, is not recoverable for the root filesystem. This it appears has to be saved as a disk image to enable a full system restoration. The operating system was reinstalled and the system configuration restored manually.
Accounting continues to be a problem and a meeting was held with John Barnes and Simon Woodcock of IBM during which it transpired that the RS6000 deals in 100 second minutes. The dissatisfaction with the current facilities was explained and subsequently some progress was made.
A new disk pack of 4x1.2GB SCSI drives was added to the system. The extra capacity is in part to be used for the ASIS software and to provide a network news facility instead of the service currently offered by Informatics . The rest will be given over to user space.
IBM have offered direct line support in return for $3000 though the figure subsequently lessens for each additional RS6000 purchased.
Fortran 90 is now available.
12.1.1.2 CAMELOT
This machine again had problems during the period some of which were solved by moving to AIX 3.2.
The card with the SCSI interface was replaced as it appeared to be causing corruption of the system disk and several reinstalls had been necessary.
12.1.2 SUN
12.1.2.1 XERXES
The XERXES Xkernel server produced one parity error during the period but otherwise remained stable.
A number of the diskless 3/50 workstations which it serves had problems making it necessary to have three monitors and a keyboard replaced. I think it is safe to say we have had our moneys worth from SUN for the maintenance we pay.
12.1.2.2 ARCU
ARCU-1 and ARCU-2 were involved in a visualisation workshop at Coseners House at the beginning of June. Both machines mount their /usr filesystem from the main ARCU server. This proved to be impossible using the default settings as the 48Kbps line between RAL and Coseners is too slow to prevent timeouts. When these values were adjusted performance proved to be practically so slow as to be useless. Does this link need to be upgraded ?
ARCU has been moved to the Cray Area to assist in the installation of the YMP8I. This has involved the installation of two extra boards to enable an FEI-3 channel connection to the XMP and the accompanying software which has involved the generation of a new Kernel. Cray have yet to make any progress with this connection and it is noticeable that ARCU has been producing less reliable since the new kernel was installed.
We still await the arrival of AIX 3.2 on UNIXFE so that this machine can run mwm for the XERXES 3/50 clients.
12.1.2.3 DIRECTORY
Formerly THOMSON this machine has undergone a full upgrade and name change following a clean reinstallation of SUNOS 4.1.1 and the directory and mail software from XTEL. Currently we are running SunNet 7.0 , Sun Coloured Books 3.1 , PP 1.5 and ISODE 8.0.3 all with the latest patches. The Rutherford DSA is stable an we are currently experimenting with the PP mailer. The details and current status of this project are covered by the TFOM committee.
12.1.2.4 NUMERICAL
The Maths Group machines are now configured to use the PP mail service provided by DIRECTORY.
12.1.3 DEC
12.1.3.1 CCDDS1
Tested METRUM helical scan tape drive, would not work except on SCSI target 5, historically this is where the tape unit is expected by the system to be.
The DECStation had a full upgrade to Ultrix 4.2A including the latest revisions which includes most noticeably Motif.
We now have the DEC OSF/1 documentation set.
12.1.3.2 CCDDS2
No problems this quarter.
12.2.1 IBM
The AIX section of Systems Group take responsibility for the UNIX front end service.
Automatic backup to the ACS utilising VTP.
12.2.2 SUN
ARCU to be moved back to Graphics area and original kernel restored.
Continue with the configuration of DIRECTORY and testing of the PP software.
Configuration of new Sparc2 Server.
Automatic backup to the ACS utilising VTP.
12.2.3 DEC
Install OSF/1 on CCDDS1.
Automatic backup of DECStations.
A new section will be included in the next report on the future of mail as this is becoming a major item.
Present: J Barlow, P Bryant, T Daniels, J Hutton
Approved.
The letter from Alister Chalmers agreeing to the new EARN subscriptions is still being prepared.
A properly signed invoice for the UK 1992 EARN subscription has now been received.
The UK is considered to have a single EARN site for subscription purposes.
Recent performance had been good. The regionalisation completion and the development of EBONE had led to files arriving in most cases within minutes. Long queues were now only building up when there was a major failure. In many cases failures were not causing problems as traffic could now be rerouted either by the underlying IP network or at the NJE level.
As there is now no intention to charge particular sections of the community for EARN services the detailed statistics are of less interest. As an experiment these will now be produced on a 3 monthly basis
EARN now produces three sets of statistics:
This was now complete. It caused a surprisingly small drop in NJE traffic as little traffic with a good Domain Address was being dealt with by RAL. On the other hand there was a steep rise in SMTP traffic as the original level was very small.
No problems were encountered in the change.
This item will be deleted from future agendas.
This is now concluded with a complete mesh of connections between core sites in place. EARN intends to close down the project.
There are IP problems with EBONE traffic leaking onto the CERN line. It is hoped that HEP will delete their direct CERN connection and use EBONE or EMPB as it evolves. This is not strictly an EARN problem.
This item will be deleted from future agendas.
As planned the German link was moved without problem to IP at the end of September. The GBOX is now redundant for EARN purposes and is being used for other purposes by CCD.
This item will be deleted from future agendas.
As the machine is no longer used for EARN purposes no further charges will accrue to JANET after September.
This item will be deleted from future agendas.
JANET is happy with the subscription levels for 1992 and 1993.
The Board of Directors meeting will be held in Pisa in November.
The UK will oppose the change in the constitution to allow the president to serve a further number of years.
The UK will oppose any budget proposals to fund RIPE, EBONE or other organisations on the basis that these fundings are a national matter.
Friday February 12 at 9am.
I have looked at 4 possibilities for a 1993 autumn EARN Conference in the UK.
Format. The conference would have to be held in a hotel and there are at least four hotels which could provide the required facilities. The cost per person would be about £146 a day. This would include hotel, conference rooms, and all food. It would be possible to just hire the conference rooms when the cost would drop to £46 a day. The hotel approached is at the higher end of the price scale. They would undertake all the organisational side of the event such as all hotel and room bookings and collection of money (we would not need Tree Emma).
Format. The conference centre is free as it is supported by the hotels. All the room we need and more is available in one building. There would be no limit on the size of the conference as we could move to larger rooms giving a capacity of 3900! There are some additional costs. Lunch £10, tea and coffee £2 per person per day. Banquet would be between £14 and £30 (average £17). Ethnic entertainment from the Morris Dancers or similar. Hotel costs are between £25 and £85 per person per night. The centre would print free booking forms for us.
Format. The event would be at Newbury Race Course which has just completed a conference facility with ideal room sizes in a pleasant location. This assumes the conference is limited to about 300 delegates. The cost would be £25 a day per delegate and this would include tea, coffee and lunch. The banquet could be accommodated. Most hotels in the area are about 50 bedrooms and in all there are about 400 odd rooms. This may be a problem. Prices tend to be £75 to £100 a night. Public transport is non-existent and we would have to operate a shuttle bus round the hotels.
Investigation are not complete (or put another way my contacts seem reluctant to grace my in-tray with any information).
Technically there is no problem with providing a 64K line. The Bournemouth option would be the most expensive as there is no convenient university near by and we would have to rent a line all the way to Rutherford. The rough cost would be £400 installation plus £200 to £800 rental. I would envisage the connection being IP and Rutherford would provide of the order of 8 PCs as terminals. We would seek support from SUN and DEC for the loan of further terminals. Rutherford would supply technical staff to set up the installation.
I have spoken to Jill Foster about a Newcastle site but she is not keen. Birmingham is a possibility as they have an international airport. As we move north to Manchester, Glasgow and Edinburgh there may be problems of remoteness for an organisational point of view. I will bring further literature to the Executive meeting.
Members: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
1 Notes of meeting held October 5, TFOM/11/92
2 Progress on mounting PP
3 Progress on purchase of a Sparc station for PP development
4 Progress on moving EARN gateway to PP
5 Progress on use of NJE under UNIX
6 Progress on study of user interfaces
SMAIL study
Studies in office automation group
7 Progress on study of alternatives to LISTSERV
8 Progress on draft strategy. TFOM/13/92
9 Progress on study of X.400 to EARN mail
10 Progress on study of address styles within PP
11 Progress with White Mail
12 Review of study areas
13 Date of next meeting
14 AOB
Members: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
Apologies P Cahill and K Hoadley, J Wheeler
Noted
Some problems have been encountered getting PP to rewrite headers. SMTP has been checked out. X.400 has not yet been tried.
Delivery expected in two weeks.
TCPMAIL cannot take the place of mailer on the IBM. Also a scheme of getting the IBM to transfer BSMTP through to PP does not look sensible.
There may be merit in moving to international domain ordering as a number of UK universities have. Can this be done and is it sensible. Action J Wheeler Action D Parker
UREP is still impossible to obtain. EARN may try and support UREP. Action P Bryant
HUJI is now on the RS6000. Tests will take place. Action T Kidd
The UNIX Andrews system which is "all singing and all dancing multi media" is being looked at. It is very large. Action K Hoadley
There are reported to be a few problems with mounting CC. Action K O'Neill
NUPOP investigations are continuing. NUPOP on the PC looks very promising. Action P Bryant
Other POP like servicers are being looked at.
The message store is still expected but is reported to be not yet stable. It will be mounted on the new PP machine. Action
The UNIX versions will be put up and evaluated. Action K Hoadley Action J Wheeler
It appears that Eric Thomas has a low opinion of these products and wants to write a new one. EARN is looking into this. Action P Bryant
INGRES is being acquired and MAILBASE will then be tested. Action R Westlake
No progress.
Awaits message store next week.
Nothing to report.
November 11, R31, 9.00.
Nothing to report.
After the extraordinary debate on EARN's development plans I have come to the conclusion that we are really wasting our time looking round for these so called "new areas". We should be building on what we have.
1. Produce standard for LITSERV. This should follow the current version rather than a development.
2. produce LISTSERV for UNIX.
3. Develop LISTSERV over IP.
4. Investigate IP version of NJE style file transfer.
5. Continue to develop conference organisation skills.
6. Continue with documentation.
7. Continue with support activities.
Dear Miss Potton,
Thanks for your help with our 1993 conference plans.
I have just returned from our 1992 conference in Pisa which was a great success. It finally attracted 360 participants.
At our Board meeting the UK bid to host the 1993 conference failed and we will hold it in Warsaw. This was not really a surprise as Poland has been pressing to host it for some time. We also decided to hold our 1994 conference in the UK and I am happy to report that from the various possible sites within the UK Bournemouth headed the list.
In the light of a 1994 autumn conference I would like to visit you early in the new year to look at the centre and discuss some of the details. I appreciate your offer of overnight accommodation but as I live only a couple of hours from you I think I will make it a day trip.
From the experience of this conference I can define our needs a little more accurately.
The conference will be for 2.5 days - starting on Tuesday morning.
We will need one hall capable of holding about 400 people.
A second hall for a parallel session holding about 200 people.
A flat floor area for 40 computer terminals. We will be providing all the computer terminals.
High speed communications to Rutherford. This will have to be provided by one of the telecommunications operators (BT or Mercury).
A demonstration room holding 100 people and 10 computer terminals.
Two or three small rooms holding 20 to 40 people for the complete week (Saturday to Friday).
We will want coffee, lunch and tea within the building the cost of which will be in the conference charge.
We will need access to the building from the Sunday to set up the computer terminals.
We will need overhead projectors in all the rooms, slide projection in one and television projection in one. We need to discuss the provision of projectors.
I will telephone you early in the new year to arrange a date for my visit.
Thank you for your help.
Yours sincerely,
Paul Bryant.
Members: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
Noted
Progress is good. There have been some setting up problems. There are a lot of tables to set up and mistakes in updating are easy to make. Documentation on updating is needed and some automation. X.400 has not yet been tried.
The Sparc station has been delivered and set up. PP is about to be mounted. 32M of memory seems tight for PP and if it is also to carry the EARN traffic more may be needed. Perhaps this should be purchased in the new year.
G Christian put forward a proposal that retains BSMTP on the IBM but reverts to standard a MAILER. Thus the BSMTP would be passed to PP in SMTP form where header conversion would take place. If this strategy is followed then IBM mail would adopt world domain order with PP converting from UK order to world order. In fact the proposal would encourage world domain order on the site. Machines directly connected to JANET via X.25 could well retain UK order but with the possible loss of VAX X.25 and Grey Book support this could be a reducing number of machines. G Christian will produce a paper. Action G Christian
With the proposal to keep BSMPT on the IBM the need for NJE under UNIX is reduced. The situation with UREP and HUJI will continue to be monitored. Action P Bryant Action T Kidd
The UNIX Andrews system is still under study. Action K Hoadley
A new Version of CC mail is expected. Action K O'Neill
NUPOP investigations are continuing. Action P Bryant
IMAP protocol will be investigated. This is more flexible than NUPOP and allows the user to keep his mail in the mail store. The user agent involved is PINE. UNIX and PC versions are on beta test. Action ?
The message store is still expected. Action ?
The UNIX versions of LISTSERV do not look satisfactory in that they lack features.
EARN is still involved in discussion to write a new LISTSERV and so to provide a standard definition. This will be monitored. Action P Bryant
INGRES will cost £1000 and thus an attempt will be made to mount MAILBASE on a PC which does have INGRES Action R Westlake
TFOM/13/92 is an interim report for TMPC. This was updated in the light of developments. Action P Bryant
TFOM/14/92 was discussed as a draft for the final report.
Awaits message store next week.
Nothing to report.
December 17, R31, 9.30.
Nothing to report.
Last year we reported the purchase of an FDDI (Fibre Distributed Data Interchange) network operating at 100mbits/sec from Network Systems Corporation. This was principally to replace the Hyperchannel between the Cray and IBM computers. The second objective was to connect a number of other machines needing a high speed network interface.
The installation was highly successful, quickly replacing the Hyperchannel and since installation has given no problems. The recently installed DEC 6000 is connected and a DEC 5000 file store service, a DEC alpha and an RS6000 are almost connected. The network has integrated easily into the existing ethernet local area network using a router which has given us a great deal of confidence in the technology.
Plans are now well in hand to introduce an FDDI network across the site. A tender exercise aimed at equipping five divisions with FDDI is nearing completion. This will enhance rather than replace the current ethernet giving the opportunity for introducing a range of new services ranging from high speed graphics to access to large file stores.
We expect to connect the FDDI network to Super-JANET so that a number of off site projects can gain access to our super computers and also to allow site based projects to access servers in other parts of the Super-JANET network.
Dear Hans,
Herewith my claim for the EARN Executive and Board meeting meeting.
Air ticket Pisa 183.40 UKL Hotel Pisa 1080000 L @ 2019.6 534.76 UKL Total 718.16 UKL
Thank you.
Best wishes
Paul Bryant.
We have completed an examination of the tenders and have the following comments:
We are happy with the technical performance of the equipment as we have experience of at Rutherford. With respect to the DECNET requirements we know that Cisco has technical agreements with DEC.
Unfortunately the price of £147429.00 is well over my target of £100000 and top limit of £125000.
We now have information that Cisco are launching a new variant of their product which has less expansion capacity but functionally is identical to the tendered equipment. The cost is about £15000 per router making a total of £75000 plus the cost of a concentrator at £7000 gives £82000.
There is one draw back in that the equipment can only support one ethernet port rather than the two required. This problem can be overcome at a cost when they develop a dual port card in the near future. The loss of an ethernet port is not serious.
DATARANGE are offering Wellfleet equipment. Their tender is difficult to understand as the exact equipment in the two proposals is not defined in detail. The cheaper offer of £121334.25 is based on fairly slow equipment and we are not happy with the performance. In addition we have similar reservations to 3COM on the DEC capabilities.
The more expensive option is way over budget.
The equipment is manufactured by 3COM. There are two mistakes in the tender. Firstly they should have included 2 3C6020 FDDI cards per router which increases the price to £83012.49. The second mistake is that they did not specify warranty which I understand is 12 months.
The principle reservation we have is that we are not convinced that the equipment will operate DECNET properly. The document states that it can in release 6 of the software which is now on beta test. The company have been unable to put us in touch with any reference site which could give us assurances. Thus, the purchase of this equipment has a high risk factor which makes me reluctant to purchase it. We are still seeking evidence on DECNET from the company.
Olivetti is offering Cisco equipment and at £237805.00 is well over budget. The salesman thinks he could reduce the cost to £194269.00 which is still way over budget.
EIT are offering 3COM equipment at a cost of £104830.00 and we can see no additional benefits. We see no advantages over Lancare.
There is a mistake in the tender in that only a single FDDI interface is included with each router and the correct price should be £171012.98. We understand that DEC have adjusted their prices but that this makes no significant difference to the price.
Technically this is the best option since experience at Rutherford can confirm that the equipment can meet all our requirements.
The price of £171012.98 is well over budget and it would be difficult to justify the additional cost. There may be some room to manoeuvre and I hope to explore some options with the supplier although I am not optimistic.
The price is a little over budget. We have reservations over their DECNET capabilities. The supplier could not support DECNET until mid 1993 but they do have an agreement with DEC whereby they will adopt DECNET router code (a similar relationship to Cisco).
The price range has been a surprise.
On paper all the suppliers can meet our needs although not necessarily on delivery (NSC for example).
We are disappointed that 3COM, as the cheapest supplier, is unable to convince us of the performance of their equipment. In addition we are disappointed that since we had preliminary discussion with suppliers they have misunderstood some points and have also come back later with alternative proposals.
If I had my way I would not purchase the equipment this year since I expect the market to stabilise and equipment to become cheaper. Unfortunately, the £100000 would disappear at the end of the year and we are expecting a tight financial situation next year. Thus my management wishes us to proceed with all speed.
In the light of events I would suggest that we shall have to go back to suppliers to ask them to confirm their tenders. (I am sure you have a better description of what I am suggesting). I suggest we speak again after you have given the matter some thought.
FDDI tender (almost complete). No action.
FDDI installation (about to start). The principle problem is the setting up of the routers. Tim Kidd has the experience but not the time. Current thoughts are that Paul Bryant, Graham Robinson and Jeremy Sharp should undertake the job with possible help from Tim Kidd.
PP investigation. Giles Christian will look after and develop PP.
Future of mail (investigation has some way to go to be followed by a number of projects). At this stage no actions are needed.
Help to Roger Westlake with UNIX.
DNS support. Roger Westlake could support the DNS although in time this should be taken over by someone such as Jeremy Sharp in operations
LAN monitoring. This could be done by Andy Jessett.
PC LPR. This will be looked after by Graham Robinson although it does not seem to be a popular product.
LLCPKT. This will be looker after by Graham Robinson. Some development is needed.
WWW. Eventually this should be undertaken by Support but in the mean time by Roger Westlake. It is a trivial job.
Operational guide to DNS. Should be completed by Kevin Hoadley.
SNMP. This needs thought.
Super Janet. It is unclear what is involved.
CLNS. A stuck project.
MACREG. The future is unclear. For the time being the service will continue. Access may be provided via WWW but it does not seem sensible to move the data base. Any WWW work will be done by Roger Westlake.
Andrews file store. As we are being encouraged by management it needs to be progressed. So far only some education has taken place. A trial is now needed. It is unclear who should drive this but a briefing is needed from Kevin Hoadley.
Kerboros. Roger Westlake should be the expert.
With loss of Kevin Hoadley we need to consider some options. The two areas concerned are the Local Area Network and UNIX. The current staff directly affected are:
Kevin Hoadley, Roger Westlake, Andy Jessett.
The tasks are:
Kevin Hoadley:
UNIX support PP investigation Some network support
Andy Jessett:
> LAN monitoring - particularly Ethernet PC network products Ethernet hardware expertise.
We can expect that we will be unlikely to replace Kevin before, say, April. In addition it is unlikely that we will get anyone of a comparable quality.
The FDDI tender no longer needs Kevin's expertise.
The FDDI installation will require router, IP, and FDDI expertise. The only sure option is Tim Kidd although it is likely that a number of people could pick up the amount of expertise needed.
PP has some way to go. Roger has enough knowledge to maintain the system but no spare time to develop it. A possible expert could be Giles particularly since the IBM MAILER has reached some sort of conclusion.
The Future of Mail, although the details are still in some doubt, the mail lines are clear. There is no option but to continue.
It is unclear just how much help Kevin gives Roger. My estimate is that we can survive without damage but progress will be a little slower. Thus, a new recruit should have some UNIX expertise.
DNS support should now become an operation responsibility. Kevin has written a useful paper on DNS support. A possible candidate for the job is Jeremy Sharp.
LAN monitoring is really IP monitoring. Andy Jessett should be able to take this on although the area needs some attention.
PC LPR may be in a satisfactory condition. Perhaps it can be frozen unless problems arise.
Dear David,
I have now discussed your requirement with my colleagues. I am happy to report that we are able to provide resources to make your data available via the World Wide Web. As you comment, I think this is the preferable service. We had some difficulty is suggesting a cost since the computing resources are small and difficult to measure and most of the cost will be in "massaging" the data into a suitable form and ensuring that it is made available. On the basis that this could absorb about two weeks of effort over a year we would like to suggest a cost of £1500.
For the fee we would be prepared to put up the data and to consult with you on how the data should be provided. We are also prepared to give you statistics on the access to the data so that you can assess the impact of the service and decide on its future after the first year. With a magnitude of some 100 papers I am happy that there are no problems with file space. Should the number grow significantly, say double, then we would need to revisit the costs as this could absorb more effort in maintenance,
Our initial thoughts are that there should be an index giving the title and perhaps key words. Each of the titles would have a separate file containing the abstract. An alternative would be to group the titles by subject and have files containing a number of abstracts. There is some work involved in preparing the files as many of the lines in a file have to be "marked" so that the information retrieval mechanisms can satisfy information searches. In the short term we would be happy to prepare the files from information you would supply on, say, a floppy disk. In the longer term you may wish to do the task yourself.
You ask whether the updating could be done from Pergamon. In principle it could as we could allow you to log into the server and edit the files. I have not given the matter a lot of thought but I think editing the files and typing the information in would be tedious and error prone. To update the files by "file transfer" would certainly be possible and we could supply suitable PC software. Clearly, if you did the job yourself you would also need to have sufficient expertise to check the results of an update.
Yours sincerely,
Paul Bryant.
In discussions with the supplier is has become apparent that the provision of phase IV to phase V translation was only present in release 6 of the 3COM software which is just entering beta test. Thus it has not been possible to obtain confidence by demonstration or discussion with a reference site that translation operates satisfactory.
As a result 10% of the purchase cost will be retained pending a satisfactory demonstration of DECNET phase IV phase V translation on the delivered equipment.
The aim of the demonstration is to ensure DECnet Phase 4 and Phase 5 traffic can be routed through the 3-COM router. It is in addition to the acceptance tests laid out in the tender document.
A DECnet phase IV VMS system will be connected to one of the ethernet ports on a 3COM router (this will probably a VAX station 3100 known as CCDVS3)
A DECnet phase V VMS system will be connected to a second ethernet port on the same 3COM router (this will probably a DEC 4000 known as RLVE running VMS and DECnet/OSI for VMS Wave 2).
A DEC phase IV system will be connected to a 3COM FDDI port on the same 3COM router (this will probably be a DEC 4000 known as COBRA running VMS and DECnet/OSI for VMS Wave 2).
Both ethernet connected machines will be in area 19.
SET HOST - A successful terminal session from each machine to the other.
COPY - A file will be copies from each machine to the other.
MAIL - DECnet mail will be sent from each machine to the other.
The ethernet Phase V machine will be moved to area >63 and the tests repeated.
The tests in 2.3 will be repeated between COBRA and CCDVS3.
All the tests should run successfully with no errors attributable to the 3COM router.
The supplier will provide assistance in setting up the router for the tests.
RAL will supply expertise for setting up the DEC systems and will provide cables.
If for any reason Rutherford is unable to provide a suitable test environment by one month after delivery the test will be deemed to have succeeded.
Dear Jim,
I have pleasure in enclosing the EARN documents which I think will be of use to you. Unfortunately some of the papers are copies from my files which have been scribbled on.
I have spoken to Hans Deckers, the EARN manager, who tells me that Azerbaijan have already applied to join EARN. I therefore suspect that things have gone ahead further and faster than we had thought at our recent meeting. Hans does not anticipate any problems with their application.
Please contact me if I can be of further assistance.
With best wishes,
Paul Bryant.
Note: This paper has not yet been agreed but is a proposal for the provision of a UNIX support service for small UNIX systems on the Rutherford Appleton Laboratory site.
The increasing number of UNIX workstations at RAL has created a requirement for a site wide UNIX workstation support service. Many users and work groups who are working with UNIX or intend to invest in a UNIX based system lack access to support services and it is at these customers that this service is principally targeted. It is worth stating immediately that the service does not include systems administration duties which would involve considerable resources but rather will offer the three phase support structure outlined below.
The following sub-sections outline the three broad areas where we offer support and advice.
The objective is to match customer requirements with available technology and to produce a system definition possibly with a set of alternative solutions.
The system definition defines in some detail the tasks that system must be capable of undertaking and defines the parameters that will be used to assess acceptable levels of performance.
The definition considers requirements such as cpu power, number of cpu's, amount of memory, amount of disk, bus architecture and bandwidth, communications and system interfaces, number of supported users and response times, graphical capability, user interfaces, applications such as database, and programming languages.
Different applications make different demands upon the system, users are advised on the configuration most suited to their needs. In addition the extra hardware and software required to allow the application to operate effectively are defined.
The nature and maturity of the operating system, added features and the availability of third party software all add to the system profile we construct. We advise on what is available and can comment on future trends.
The cost of a system is in some cases more important than its precise specification. We make it our business to hold all the current information from a broad range of suppliers in areas such as pricing strategy, discounts, special offers, all of which contribute to machine selection.
We offer help and advice on all aspects of purchasing and installation of UNIX systems such as:
- ordering
Constructing the order, ensuring part codes are correct, ordering from suppliers and Value Added Retailers. Help and advise on further applications and upgrades following the order.
- inspection
Ensuring that all units are delivered, checking components are in satisfactory condition, dealing with stores and getting the system to the user.
- configuration
Installing the hardware, the operating system, configuring it for use, networking and addressing, 3rd party hardware and software installation.
- maintenance
Advice on what maintenance deals are available, what should be maintained, software upgrades, 3rd party hardware and software maintenance.
In addition to the acquisition and initial configuration of a UNIX system we provide a general level of support to new and existing UNIX users. We offer help in areas such as;
- general queries
We offer general help, advice and information on products, machines, and operating systems. This support is expected to be wide ranging and where possible in depth investigations will be carried out, with the assistance of VARs and manufacturers if necessary.
- product evaluation
We monitor the market place for useful applications free or otherwise. We keep the users informed with a newsletter.
- fault reporting
As part of maintenance and keeping kit serviceable we deal with the engineers who fix problems and make sure that they do a satisfactory job.
We offer all of the above for the following platforms;
IBM RS6000
SUN 3/x , 4/x , SPARC
DECstations, Alpha AXP
SGI Indigo
Further platforms will be added as they become popular.
We advise on the following UNIX flavours;
SunOS Solaris OSF/1 Ultrix AIX Irix
The UNIX operating system has many variants and these in turn are modified by the system vendors to produce their own particular environments. Initially the machine range supported must be restricted but it is envisaged that this will be expanded in future.
We are however in a position to provide support for all the major UNIX variants and as such we feel this is a useful service to offer to the RAL UNIX community.
Support can be provided in two forms:
a) By contract
Divisions or groups can purchase support at a rate of 1/5 of a man per machine. Machines purchased during the year will be supported free of charge on the expectation that they will be financed in the subsequent year.
b) On time and materials
Support will be provided on demand for unsupported machines and will be charged at the normal rate. This support will not take precedence over support provided by contract.
Trivial enquiries will be answered free of charge.
Rutherford Appleton Laboratory PC Support charges 0.01my per PC per financial year for its services. This figure is calculated annually around the end of the year and is based on the number of PCs registered with PC Support plus orders that are going through the system.
A PC becomes registered by either being bought through PC Support or may be added to the list at the request of its owner at the start of the financial year for which payment is received for it.
PCs purchased through PC Support receive free support for the remainder of the financial year in which they were bought on the assumption that a charge can be levied on them in the next financial year.
PROFS PCs, that is those PC systems recommended, installed and supported by the PROFS team, are also registered with PC Support but are not charged for directly. They receive a similar service except that their first line support is via the PROFS team.
No support is given for PCs not registered with PC Support and management reserve the right to curtail the general support given to any group which has a significant number of unsupported PCs. Any attempt to bend the rules by moving a problem from an unsupported to a supported PC will severely impair the goodwill towards the group involved. We will endeavour to play fair with you if you play fair with us.
Notwithstanding the above, management reserves the right to curtail excessive demands on the service especially during periods when RAL management does not supply the necessary manpower or resources.
A telephone support service and, at RAL only, on site visits where needed, to sort out PC problems. This will usually be within a working day depending on the urgency. Such a service from a Dealer costs a minimum of 30 pounds per hour.
A maintenance service that aims to get you going again within two working days for the price of parts and external repairs costs. A stock of spares including monitors, keyboards, power supplies, disc drives and interface cards is maintained. Once a fault has been diagnosed then the failing item is replaced and repaired at leisure. A limited number of complete units are kept for temporary loan when a need is critical.
A PC hardware advice service which recommends PCs that have already been evaluated for their suitability at RAL with special emphasis on communications (3270 and Ethernet). The aim is to purchase machines which can have a wide range of uses thus giving them a long working life. The recommendations are under constant review in the light of changing needs and prices.
A PC software advice service which evaluates software packages that are likely to be of interest to RAL and maintains an awareness of what is available. A limited amount of research to find suitable products to meet specific user requirements is undertaken. Site licences are purchased where appropriate. PC Support also acts as a contact point for RAL users to find out who else is using a PC for similar work.
A PC purchasing service which orders and installs PC hardware and software, both for the initial system and upgrades. This includes taking care of all the administration involved. Both software and hardware are purchased at educational discounts and this is usually at a considerable saving on normal discount prices. The average staff cost per PC purchased is one man day.
The availability of demonstration systems showing recommended hardware and software. This includes large screens, networking and current releases of the recommended software for word processing, desk top publishing, spreadsheets, databases etc.
A disc recovery service for both floppy and hard discs. This service would cost about 450 pounds per hard disc from an outside supplier.
A virus removal service including provision of an Antivirus program to detect a virus attack. Expertise on the latest developments in virus'.
Free provision and maintainance of software to meet RAL requirements that cannot be obtained economically from elsewhere. These include the MOS 3270 Emulator, an Antivirus program, a Device Driver and TSR loader/unloader (Loadsys), and the RAL utilities which include a simple to use backup system.
Installation and support of TCP/IP networking software for PCs, a lot of which is free or covered by a site licence. The TCP/IP software includes Network File Services to a remote server (NFS), X_Windows, an IBM full screen terminal emulation (TN3270) and a remote print server (LPD).
A news letter with product reviews, cautionary tales and general advice about twice a year.
And finally access to a group of people whose main interest is the PC on which they have a considerable amount of expertise. They spend a significant amount of time keeping that expertise up to date and expanding it. This includes reading the industry literature and noting market trends.
Computer network services have become all pervasive on the Rutherford Appleton Laboratory site and within the world wide academic community. Developments are continually being made both in hardware and network protocols. These developments offer new or faster services which customers may wish to take advantage of.
It is important that services provided are coordinated with the JANET services and also across the site so that connectivity can be maintained and improved.
The services provided are:
To provide a reliable and performant service and an economic cost whilst taking account of the customers needs requires the development of strategies in various areas. For example, a strategy for the development of the local area network has been developed and is now under implementation, a strategy for the development of electronic mail is under development.
Contact is maintained with customers, suppliers and the JNT is maintained. A number of coordination meetings are run. The current list includes IP, mail and the development of the local area network. Various meetings held by divisions and other organisations are attended and support offered where it is in the interests of the customers.
Various possible network developments are researched with a view to advising on their applicability to the Laboratory. This ranges from discussions with suppliers and paper studies to the evaluation of equipment. Currently connectionless OSI services, network monitoring, and various mail products are being looked at.
Some software is developed where suppliers cannot meet the Laboratories needs. This is mainly in the PC and the network monitoring areas.
Almost every computer system has a network connection. It is therefore appropriate to centrally fund the activity since it benefits most customers.
It is important to discourage groups setting up their own network services which may go against the interests of the Laboratory. Funding on an individual customer or group basis would encourage such activities whereas central funding discourages it.
Members: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
1 Notes of meeting held November 11, TFOM/15/92
2 Progress on mounting PP
3 Progress on moving EARN gateway to PP
4 Progress on use of NJE under UNIX
5 Progress on study of user interfaces
SMAIL study
Studies in office automation group
6 Progress on study of alternatives to LISTSERV
7 Progress on draft strategy.
8 Progress on study of X.400 to EARN mail
9 Progress with White Mail
10 Review of study areas
11 Date of next meeting
12 AOB
The Microcom Bridge/Router is a bridge/router with the advantage that it can compress data on a serial line. It is a candidate for improving the utilisation of the RAL/CERN line.
The equipment is an PC/AT with an ethernet board and a serial board. The ethernet board has a 68000 processor which undertakes the routing and bridging. The serial card has a 68000 processor which is used for the compression. The serial card can use HDLC or X.25.
The compression can achieve a compression ratio of 4:1 at speeds of up to 512K although the board will work up to 2M. Above 512K the processor cannot keep up with the line and the compression falls off to about 1.5:1 at 1.5M.
In practice the compression achieved can vary widely. On some binary load images almost no compression is possible while on some graphics files it may be as high as 8:1.
The PC/AT bus is sufficiently fast to service the boards.
For the CERN connection the easiest and safest scheme is to use the Microcom equipment in bridging mode and to retain the Cisco routers at each end. In effect this would make the use of the equipment transparent. With such transparency there is less likelihood of CERN or JNT objections and a high level of confidence of success.
Unfortunately, an additional ethernet port on the JIPS Cisco router may be required and possibly an additional port at the CERN end. If HEP provide their own router this would need two ethernet ports and no serial port (to CERN).
A more risky option is to use the equipment in routing mode. In this case it would connect to the CCD ethernet at the RAL end and a similar network at CERN. There are two problems. The first is that the exact quality of the Microcom routing code is unknown and there would be a risk of this upsetting the networks at each end. Clearly, with suitable testing this risk, albeit small, could be minimised. The second problem is that it would be yet another type of routing equipment to set up and maintain. In addition, CERN and the JNT may not be happy with such routing equipment.
There are an abundance of possibilities for improving the performance of the CERN line.
In principle IP, DECnet IV, and DECnet V could be routed by many of our routers and could use a single HDLC connection to CERN via a Microcom. It should be noted that the DEMSA could not be used as it is not capable of supporting more than one ethernet port.
Some possibilities that could be considered include:
A similar range of possibilities exist at CERN.
It is possible to switch on header compression in the Cisco routers. This reduces the length of an IP header from 32 to 4 bytes but has an appreciable overhead in the processing needed in the router. It has a very beneficial response effect with telnet traffic which is characterised by very short packets. It has little effect on bulk traffic which is characterised by large packets.
The Microcom costs are:
Wide area board £2555 Ethernet board £1050 Software £ 445 PC/AT £1050 Total £5100
The PC is as recommended by the supplier and is a rack mounted very reliable model. It has 512K of memory and a floppy disk keyboard and display. An alternative from our normal suppliers would be £600. This would have the reliability of normal desk top PCs.
Possible additional costs:
If it is intended to continue to use a Cisco for the routing then an ethernet port is required into the JIPs router or the expected HEP Cisco router.
If the JIPs router is to be used this will have to negotiated with JIPs management and may require finance for an ethernet port which cost £6420 per pair but subject to substantial discount.
If the expected HEP Cisco is to be used the cost will depend on the model purchased.
An ethernet board for the DECnis at about £1000.
The 3COM equipment will be delivered with spare ethernet boards which cost £621.
Similar equipment will be needed at CERN.
Thus the best cost is £9520 and the worst is £25040.
The CERN line costs £150K per year or £586 per K of bandwidth per year. The use of the compressor on the 57.6K IP connection would, with 4:1 compression, generate an extra 172.8K which would have cost £101K per year. If used with the complete 195.2K of the IP, DECnet IV and DECnet V the extra bandwidth generated would be 586.6K or £343K a year.
This is a rather optimistic calculation! In practice I would expect something like a 2:1 compression. The calculation makes bold assumptions that the 4:1 compression is achieved. It assumes a linear cost for bandwidth. However, it does suggest that for an expenditure of £9.5K and £25K increased bandwidth to the value of £343K a year could be achieved.
No doubt the reader will wish to do his own creative economic calculation!
Microcom operate through agents. Eurostar Network Systems are prepared to sell us the equipment on the basis that it can be returned within 30 days if found not to be suitable. This would allow tests to be carried out. It would be difficult to undertake a "laboratory" test as the nature of the CERN IP traffic is unknown with respect to its compressibility. The only realistic test would be to take equipment to CERN and conduct a test with the live traffic. This would still require the possible purchase of Cisco and possibly DEC equipment if the bridge option was to be taken. It would require some organisation to ensure that the tests could be completed within the 30 days. None the less it would be possible as long as manpower can be found.
The second company which market the equipment is BTN who visited me but who recommend we purchase the "reliable" PC from them.
The basic costs of the boards is the same from each company but discount may be possible.
I have spoken to Martin Jordan of Oracle who use older versions of the Microcom equipment on a 19.2K line to Scandinavia. They use it as a bridge with Wellfleet and Cisco routers inboard. The traffic consists includes IP and the two flavours of DECnet.
Since installation over a year ago there have been no problems or failures attributable to the Microcom equipment.
Oracle claim to obtain a 4:1 compression on binary files and a 2:1 compression on text.
I have made some preliminary enquiries with CERN. In principle there does not seem to be a problem. In practice there may be problems with equipment limitations but these would need examination if and when a definitive proposal is produced. Of course finance would be required if any upgrading of CERN equipment is involved.
The various network projects - including DECnet phase V, JIPs, the updating of the local area network, Super-JANET as well as the Microcom equipment - suggest that we should be having a close look at the possibilities that are now opening up for providing a better service more economically to the HEP community. I have no doubt that we could pick one of the options above and make it work it does beg the question as to where it would lead us.
Members: P Bryant, P Cahill, G Christian, P Girard, K Hoadley, T Kidd, K O'Neill, D Parker, R Westlake, J Wheeler
Noted
X.400 will progress when it is clear that there is a real need. It requires registrations in the NRS which needs investigation.
The PEB problem on domain reversal remains and has been referred to X-TEL.
Assuming the PEB problem can be resolved the target date for service is end of March. At this stage only the relay service will be offered. The name server, message store and message store will follow.
Currently there are about 24 test users.
J Wheeler will look into the publicity required for service. Action J Wheeler
The service will be provided on LETTERBOX and any future testing will be on DIRECTORY.
There may be a need for more memory on LETTERBOX towards the end of 1993.
There is quite a bit of testing to do before the "Christian Plan" will be complete. The target for the tests is the end of January. However, the scheme will not be brought into use until LETTERBOX has been in use for enough time to gain confidence in it.
G Christian will circulate a copy of his plan. Action G Christian
Mail has been received from one who claims responsibility for UREP. This claim is being followed up by T Kidd. Action T Kidd
CC has now been loaded. There is some concern as to whether its X.400 gateway provides the 84 or 88 version of X.400. The evaluation should be complete by the end of January.
A similar trial will be carried out with MS mail. Action K O'Neill
INGRES is being ordered for the SUN after which MAILBASE will be evaluated. Action R Westlake
No progress. A final report should be ready for the end of January. Action P Bryant
Still awaits message store next week.
Nothing to report.
January 28, R31, 9.30.
Nothing to report.
Hans - I have drafted a letter for Bob. The letter in long but says very little. In the main it tells Bob to go away and read some papers. There are a few points that I have not checked. Did we send the last BOD papers to James Hutton? Have we always sent a set either to Bob or James? If not amend the letter. I am not sure where the monthly updating of the routing tables is defined to which this letter refers - this needs checking or deleting. There may well be other details you might want to change. The main rule is to refer him to papers and not to send them to him - this will keep him busy!
Dear Bob,
Thank you for your letter requesting further information on EARN.
The way in which the UK subscription was arrived at is documented in the Board of Directors paper BOD25 92 which defines the computation mechanism. The actual figures used in the computation are in BOD38 92. From this you will see that the UK is regarded as having one EARN site. I have pleasure in attaching the relevant BOD papers. A set of the Board paper were sent to James Hutton. There are all available electronically from the LISTSERV at FRORS13.
You asked about the services provided by EARN. The way EARN operates is defined in the EARN statutes. From time to time EARN has produced a series of additional papers which further define it operations. These documents, which are all Board of Directors Papers, include:-
EARN Charter (BOD....) EARN Code of Conduct EARN statistics directives Operational procedures for connecting a new country Country payment procedures LISTEARN "Licence Agreement" EARN SNA directives and recommendations EARN operational procedures Line sharing procedures Management of change procedures for EARN. RPG proposal for regionalisation of EARN.
All significant decisions with respect to EARN result in a Board papers for a Board decision. There may well be other papers of interest to you. A full index is distributed with Board papers and is also obtainable electronically as EARN INDEX from the same LISTSERV. In some cases papers have been rejected by the Board and the relevant Board minutes should be consulted for such information.
The technical control of EARN is mainly devolved to the NOG (Network Operations Group) which has defined at the detailed level the way the network is operated and the fine detail of the service levels. Many of the details of their work are defined in the minutes of NOG meetings which are held twice a year. The minutes are published as Board papers. These deal, for example, with the details of the monthly updating of the routing tables which places obligation on EARN (through its staff which operate the procedures) and on sites who have to install new tables.
You will no doubt be aware that EARN has been involved with several activities at the direction of the Board. The exact nature of these is well documented in Board papers. The resources absorbed by these activities is documented in the Board papers which show in fine detail how staff spend their time. In addition, other papers show the progress achieved.
You will no doubt be aware that EARN provides its services and operates its projects under the direction of the Board of Directors. These services and projects are co-operations between the EARN staff and volunteer staff at various sites. Services can be impacted by events outside of EARN's direct control such as line failures, the EBONE and sites themselves. Thus the exact level of service is not entirely under EARN's control. However, in the papers you will see the service aims of EARN. These service aims are continually changing as technology improves and as the means of measuring service develop. EARN now monitors performance carefully and service levels achieved are monitored and compared with historic achievements. Where there has been a degradation there are procedures for investigation and rectification.
EARN is happy to receive payments in ECUs and looks forward to an early payment.
Finally I can confirm that the legal entity is the European Academic Research Network. The details of the foundation of EARN are to be found in the EARN statutes which have been referred to above.
I hope these notes have answered your concerns but please contact me again should there be any further questions.
Yours sincerely.
In many areas of life customers are expecting some form of agreement on the services and the service performance levels a customer may expect. This allows an organisation to evaluate the cost benefits of a service as an aid to deciding whether to subscribe and as a way of measuring performance.
EARN, as a service provider, should be in a position to provide such a service definition.
In defining the service a distinction needs to be drawn between development activities and those activities which directly affect the customer. This paper address the latter activities.
The items in the service definition below are illustrative and members are invited to comment on the advantages of producing such a definition and the form it should take.
The services EARN provides are:-
During the last few months the testing of DECnet phase V has progressed well. There is little doubt that a service world wide for the community (particularly HEP) will develop. The emerging HEP IP and phase V services across the UK and to CERN are proving them with a good range of services.
A concern is where the combined HEP and community networks are leading us. On the one hand we see the exploitation of Super-JANET and EBONE to provide high band width services available to all and on the other we see a parallel HEP IP service using the current JANET X.25 network and the CERN line. We also see a phase V service about to emerge with no JANET equivalent. This raises the question of how the best services at the cheapest price should be provided.
2.1.1 Coloured Book Services (JTMP, NIFTP, SSMP)
Stable apart from occasional JTMP problems.
2.1.2 Network/3270 and line-mode terminal services
Some re-working of the Network/3270 support in VMNCP was needed because VM/ESA was found to be assigning logical devices by an undocumented new algorithm. This had been causing a number of calls to hang. The problem is now solved.
2.1.3 Interlink DEC/IBM gateway
The intermittent delay problem reported by R1 users turned out to be unrelated to the Interlink software. It was caused by a defect in some equipment on the Ethernet, and has now been fixed.
No other problems have been reported, and this service can now be considered stable.
2.1.4 VTAM
Stable. SSP V3R7 has been installed.
2.1.5 IBM 3745
The latest software releases seem to be reliable. There have been no abends or other unscheduled outages this quarter.
2.1.6 Configuration
No significant changes.
2.1.7 X.25 and Pink Book
VMNCP 7.6 was finally released at the end of the quarter after several weeks' delay because calls in the Netcomm PSE appeared to be hanging. This was finally traced to an unsuitable timeout parameter in VMNCP, and has now been fixed.
An attempt was made to raise the packet size above 256 bytes. Unfortunately other sites could not "negotiate down". No further attempts are planned.
2.1.8 RSCS
We continue to run Version 2.3.05. We plan to install Version 3.1, but progress is slow due to other work.
2.1.9 TCP/IP (including the IBM 8232 and the NSC 7000)
The NSC Ethernet/FDDI router (Gaius) experienced its first fault. The router halted, but a restart worked and the fault has not returned.
The VAX/6000-620 has been connected to the DEC FDDI concentrator.
GBGBOX has been removed from the VMNET (RSCS NJE over TCP/IP) service as the UKACRL node is now fully "mesh" connected to EARN nodes via VMNET directly from the IBM 3090.
2.1.10 Other IP Based Projects
The Virtual Tape Protocol use is increasing. Work has started on a VTP "front end" for the IBM 3090 following the delivery of the RS/6000-340. Jeremy Sharp (from Telecoms Group) is working on a VTP server for Unix machines.
The TAPE command continues to be enhanced, driven mostly by pressure from users of the new VAX/6000 service.
The work on an experimental network calendar server which provides access to the OV/VM calendar engine over an IP socket-based interface has "stalled" due to pressure of other work.
Some work has been completed towards automating the generation of mapping tables for the Cray NFS.
2.1.11 MAILER : Present configuration
Two bugs have been fixed on the JANET side of the gateway , and one on the EARN side.
Non-reversible From: addresses are now only rejected by EARNGATE if the mail is destined to go to MAILER.
Mail from the new Cray can now be replied to.
Apart from the above the email service has been stable, and the problems caused by domain reversal, raised by User Support, have been shown to be caused by other sites.
2.1.12 MAILER : The future
A first-cut design for a replacement of MAILER and EARNGATE has been produced. The requirements of the design for the system have been stated.
2.1.13 RS/6000
The new version of Coloured Book software needed for AIX 3.2 eventually arrived, and the X.25 and Coloured Book support is now ready for use as soon as AIX 3.2 is itself ready. Meanwhile some existing problems have been solved by regenerating the NRS tables under 3.1.5.
The new RS/6000-340 has been installed, but despite the installation of many tapes from IBM (and at least four versions of the operating system) FDDI still fails to work. IBM are sending AIX V3.2.3 which is expected to reach us early in December, this may cure the problem. In any case it shows that we will still be unable to run FDDI on the 550 with the AIX 3.2 being tried at the moment.
2.2.1 Migration to the VM/ESA version of GCS is likely.
2.2.2 Investigation will continue into Unix-based JTMP software that could replace the VM-based JTMP server.
2.2.3 There is a need to bring all the VTAM and NCP related software up to date by applying the latest PUTs and PTFs.
2.2.4 Installation of RSCS V3.1 on the IBM 3090.
2.2.5 Installation of TCP/IP V2.2 on the IBM 3090.
2.2.6 The updating of NFS mapping tables on the Cray should be made automatic.
2.2.7 A test version of the new email configuration will be in place on the IBM and Sun.
* R18 have agreed to go ahead with the floor replacement and now require details of how and when. As the end of the financial year brings its normal rush of activities (SuperJANET and FDDI etc) I have decided to leave the start of the replacement until the new financial year. This should give us enough time to plan what's going where during the upheaval.
* The early part of the quarter equipment damaged during the storms in August was tested out and either repaired locally or sent away for repair. This included 4 power supplies and 5 channel boards for Camtec PADs. Linedrivers were not repaired.
* Two power supplies and two channel boards were also replaced on fibre optical multiplexors.
* The CPSE has been reliable, the traffic level on this appears to be dropping.
* The SEEL had two problems during the quarter. The number of LCNs on the JANET link had to be increased from 96 to 128 due the increase in the number of calls. A hardware problem arose during the air-conditioning shutdown - after the CP95 test, the switch would not reload, the battery backed up RAM was lost and attempts to reload from disc failed. The system was reloaded from disc after the air-conditioning was restored. Subsequent tests proved that the disc controller was flaky and duly replaced.
* The Multiplexed Megastream line to Swindon supporting the Vtam-Vtam, DBS and Lee Data links had three outages. These followed an upgrade of the multiplexors to support speeds greater than 1M on a channel. The mux timing was set to be taken from channel 1 and not internal - this was believed to be the cause. However the problem was still evident after the clock was set to internal. Tests on the line and the mux indicated that all was OK, however they would not work when connected together. Further tests by BT indicated that the line was suspect to 'All 1s'. When we retested the line with this pattern it also failed. BT located two faulty boards at an exchange near Swindon. Consideration is being given to some form of backup for DBS work in an emergency.
* Poor response on access to the new VAX on the 'Services' ethernet led to a good deal of work trying to pin the problem down without causing too much disruption to the normal service. The problem was eventually attributed to a port on a Fibre Optic Multi Port Repeater (FOMPR). The way the FOMPRs are connected together and the distribution of the Villages between the FOMPRs has been changed following this incident. This has also resolved the problem with the Interlink DEC/IBM gateway, see last quarterly report section 2.1.3.
* The outage of the JIPS router for an upgrade affected routing between the two RAL IP networks. Users have been advised to use GAIUS as the preferred gateway.
* Both International lines had a single outage, but not connected.
3.1.2 JANET
* The traffic levels have increased dramatically, with an average weekday throughput of 8+ gigabytes on the Atlas2 switch.
* A considerable increase in the JIPS traffic has caused problems with the loading of the CPU on the router. The line speed has been reduced to lower the demand, as a short term solution. It is intended to install an additional board and then run native IP between RAL and ULCC routers over a 2M circuit - this is scheduled for 10 December.
* The Netcomm switches have been generally reliable, unfortunately all the non-scheduled breaks were on the main switch - 3 in total. One due to a Self-Restart with no explanation, another due to a high number of Internal DXM errors which caused user problems over a weekend and once when the system was power cycled due to at least two ports and the manager port dying.
* Mercury lines also were generally well behaved. Two points of note: once when all lines were lost for a couple of minutes, Mercury suggested weather conditions; the other was when the Swindon link was de-programmed.
* BT also seemed to be fairly quiet. A break in a High Order route outside of Oxford took half our circuits down for a few hours, there were other minor outages on digital circuits and three on analogue.
* An improved version of the Network Management System has been installed for the operators.
* JIPS Router upgrade.
* Installation of FDDI cabling and equipment.
* Installation of SUPER-JANET equipment.
* Take responsibility for the CCD FDDI.
* FDDI monitoring equipment.
There were 89 reported terminal faults in this period. Surprisingly there were still a few terminals reported that were damaged during the storms of August.
The sub-section staff repaired 11% of the faults, the remainder were done by maintenance agency.
The distribution of agency repairs is as follows:
Last qtr. This qtr.
DCM (Kode) 75 79 Contracted terminals
OTHER 0 0 Warranty repairs
The period's faults have been analysed for divisional distribution. Of this total, 24 (27%) were CCD and the remaining 73% for other RAL departments.
_______________________________________
| | REPAIR AGENCY | |
|Division |____________________| |
|-Project | TCOM | DCM | OTHER| TOTALS |
|_________|______|______|______|________|
| ADM | 1 | 5 | | 6 |
| BNSC | | | | |
| CCD | 4 | 20 | | 24 |
| CO | | | | |
| EBW | | | | |
| INF | | 15 | | 15 |
| PPD | 2 | 3 | | 5 |
| RCR | | | | |
| SCI | | 13 | | 13 |
| SL | | | | |
| SPA | 1 | 13 | | 14 |
| TEC | 2 | 10 | | 12 |
|_________|______|______|______|________|
| TOTALS | 10 | 79 | | 89 |
|_________|______|______|______|________|
This is the first quarter where DCM management was in control. During the change-over period one problem was encountered. The Atlas HP Laserjet IIISi required 200,000 copy maintenance kit, which was duly ordered via DCM and took so long to arrive that it had to be reordered. The reorder took only 2 days to arrive. It was later found the original had been sent to the wrong address. The delays took the time-to-repair over the contracted 14 days, where on-site repairs don't usually exceed 2 days. The second round of 1992 preventative maintenances on laser printers has begun and should finish in January. There is now 106 printers on this schedule.
Establishment of a maintenance contract for the new HP LaserJet IIISi and colour system printers. DCM is being asked to quote.
Several coaxes were installed for '3270' devices.
Coax, PAD and DECserver connections were changed around to accommodate office changes around the site.
Thin ether connections were installed in R2, R12, R18, R25 & R65.
Appletalk cabling was installed in R12 and R18.
A thick ether cable was installed between R12 & R65, two thick ether cables were installed between R25 & R68 and one thick ether cable installed between the North and West computer rooms at the Atlas Centre.
Data cables in the remnants of R20 were routed to R68 to maintain services.
A Camtec PAD was installed in the Technology Electronics Lab. (ex R20) in R12.
FDDI cables were installed in the Atlas Centre.
Wiring was installed in the Directorate to support a new printer.
A problem involving devices on the Services Village was investigated and identified a faulty FOMPR port as the cause.
Wiring will be installed in R1, R18, R25, R27 and R30 for new printers.
DECservers will be installed in R1, R2 and Telecomms to support those printers.
A fibre optic cable will be installed between the MRC and Atlas Centre to link their ethernet to ours.
Fibre optic cables will be installed from R1 to R55, R1 to R25, Atlas to R1 and Atlas to R55 to provide an FDDI service.
Various wiring changes may be done because of EBLF moving into the current Informatics computer room in R1 and Informatics moving their Computer room equipment into Lab 11.
Thin ether cabling will be installed in R3, R31, R18, R65, R66 & R68.
A new rack may be installed in R65 to house the expanding comms equipment that supports R12, R63, R65 & R66.
Much of the activities happened in the past quarter relate to the VAX 6000/620 and the Alpha VAX DEC 7000/620 which were delivered in late September and mid-November respectively. A user service with full networking access was introduced on the VAX 6000 soon after its arrival.
On the DEC 7000 front, there are some delays as we are still waiting for some hardware (second CPU, FDDI card and some DSSI disks) and software from the manufacturer. Nevertheless, a minimal operation system with DECnet and LAT has been installed on this system.
The UK HEP/SPAN DECnet/OSI pilot project has made some good progress in this quarter. Both Manchester, Oxford and Cambridge have some systems running the DECnet/OSI for VMS (Wave 2) and they have all established connectivity with Rutherford and the latter's central name server. Three Network Integration Servers (NIS) boxes borrowed from Digital have both arrived at the three sites. The next phase will be to set up Phase 5 traffic across using these NIS boxes and dedicated X.25 lines.
6.1.1 Network Software
There had been no software updates on PSI, LLC2, CBS and JTMP.
An update to TCP/IP for VMS (UCX V2.0) has been released. This update provides FDDI device support, which is particular useful for the CCD VAX 6000/620 to enable data transfer with the storage devices using the Virtual Tape Protocol. It also provides support to SMTP mail as well as other services such as LPR and remote shell commands.
Several updates to the DECnet/OSI for VMS field test kit were received and tested on the CCD Development VAX. Yet, there are still some outstanding problems with the software with special relevance to the HEP/SPAN DECnet environment. They have been fed back to Digital engineering group.
The monitoring of active DECnet hosts and connections and the registration of the machines in the RAL DECnet area (Area 19) continued. Currently there are 857 active machines out of a limit of 1023 allowed and 443 of them have provided us their networking details. The information collected will be useful for planning the DECnet/OSI migration for the UK HEP/SPAN community.
From October 6, the GBGBOX link to DEARN (Germany) link has been stopped. All EARN traffic are now being routed through the VMNET on the IBM.
6.1.2 Network Hardware
Soon after the arrival of the VAX 6000/620 in the Atlas Centre, DECnet, LAT, UCX (TCP/IP) and PSI Access (X.25) were installed onto the system. As mentioned above, one can exchange data with the Cray and the IBM directly using TCP/IP (UCX) over FDDI.
On the DEC 7000/620 which was delivered in mid-November, DECnet (Phase 4) and LAT have been installed and both are running over Ethernet. Other layered software are still not available from the manufacturer at this point in time.
A compact disc based information server, Digital Info Server 150 has been installed on the Service LAN in early November. This system consists of five CD drives, each of them can be loaded with a Digital media or documentation CD and accessed by all VMS machines across the site Ethernet.
No changes in the network hardware on the other VAX systems in CCD were made in the past quarter.
The regular networking support of PSI with PSI Access , CBS and Red Book, Pink Book, DECNet, LAVC, PSS/IPSS, VMS-COMMS, EARN and Janet mailshr continues. The NRS updates have been continually supplied to SERC sites.
A maintenance contract has been agreed in principle with Digital to provide maintenance to the Digital equipment on site for the next three years. This also includes a DECcampus agreement, which allow users under the maintenance to use any Digital software under pre-defined conditions. A final contract agreement will be drawn up and forwarded to the Rutherford Contract Group for their final signature of approval.
No major security problems have been recorded in this quarter.
Major efforts will be spent on the DEC 7000, pending on the arrival of the necessary software and hardware, to bring out a full production as soon as possible.
Digital is still in the process of negotiating with UWCC to renew the CBS contract, which has been described by Digital to be "close" to completion. When that is in place, the VAX group will try to participate in the field test of the next CBS release which will provide the VAX clustering support.
The field test on the NIS 600 router and DECnet/OSI for VMS (Wave 2) will continue. Work will begin to set up Phase 5 circuits across JANET using the specially provided NIS router and X.25 lines, between Manchester, Cambridge, Oxford and RAL.
The DECnet nodes monitoring and registration will continue.
7.1.1 NJE
All EARN traffic is now via VMNET and IP. There is a complete set of connections to the other EARN core sites plus a link to the States. See section 2.1.9 and 6.1.1. This has eliminated the criticism of the UK by EARN on this topic.
The traffic levels are stable although from day to day there are wide variations.
Performance, as seen from EARN's monitoring, is very satisfactory. Figures are collected on queue lengths and file flight times. Sites are asked to explain any drops in availability below 80% in any month. This quarter there have been no such requests.
7.1.2 File Transfer Between EARN and JANET
No change.
7.1.3 Mail
See section 2.1.11 and 2.1.12.
7.1.4 LISTSERV
The problems with updating LISTSERV have been solved and LISTSERV is now up to date. The problem was thought to be due to local VM modifications which prevented a disc being accessed,
7.1.5 EARN Management
At the EARN Board of Directors meeting the budget for 1993 was agreed. This has not changed significantly from the draft budget agreed in the Spring. Thus the low subscription for the UK is maintained.
The most contentious issue was the future of EARN where the Executive was criticised for not coming forward with a proposal or more accurately, an acceptable proposal. From the discussion it was clear that the views on EARN's future are varied. They range from the phasing out of NJE and EARN, through supporting new services (although no one had much idea what they were) to moving into support and documentation areas. This discussion has some way to go although the Board is expecting some more proposals at the Spring meeting. This meeting will be held in Trondheim in conjunction with the RARE conference.
The Pisa network Users Conference was a success. It clearly fills an important gap since the RARE conference tends to concentrate on technology.
7.2.1 NJE
RSCS version 3.
Further investigations into possible NJE products for UNIX.
7.2.2 File Transfer Between EARN and JANET
No change expected.
7.2.3 Mail
Further work on moving the gateway to PP.
7.2.4 LISTSERV
Further examination of LISTSERV under UNIX and MAILBASE as an alternative to LISTSERV.
7.2.5 EARN Management
No change expected.
The 1993 Network Users Conference will be in Autumn in Warsaw. The 1994 conference will be in Bournemouth.
The statistics for PC orders over the last three months are 227 orders placed, which included 68 PCs, at a total value of £218,714. Up to the middle of the quarter PC orders were at the same rate as previously but during November the order rate doubled giving a net increase of 54%. The grand total of PCs is now 815, up 212 on last year, which is yet another record for the annual increase. The normal monthly PC order total has already been reached in the first two days of the next quarter so the rush is not yet over.
In consequence PC Support has spent a large amount of its time just placing orders, and this has not been helped by rapid fluctuations in prices due to the sinking pound and worldwide price increases for essential components such as memory and Intel 486 chips. The general price trend is now sharply upward. A move into a new room for assembling and testing PCs has just been completed in time as the orders are starting to arrive. Despite the price changes Viglens continue to be very good value for money. It was salutary to note that another PC company, Ti'Ko, recommended by a user and briefly looked at by PC Support, is now on the verge of receivership.
The process to recruit a new member for PC Support, which was started in June, has now got to the stage that an advert is being drafted. At the rate it is going the number of PCs in the lab will have risen enough to justify a second post by the time interviews take place.
The steady stream of breakdowns has continued with PS/2's again being the most frequent culprit. PC Support has been a test user for the new Service Line IR system whereby IRs arrive electronically, and the results so far have been encouraging.
Recent developments in fast graphics cards have made it worthwhile to re-evaluate a PC as an X terminal. A PC has therefore been equipped with a 21" screen and a fast graphics card at a similar cost to an X terminal. The results are very encouraging with the response and resolution being pronounced most acceptable by the user of a real X terminal. Although the PC was slower in extreme cases for anyone other than a serious user of X it would make a more versatile alternative. X benchmarks so far give a performance of 31,000 Xstones and further tuning could well improve on this.
The steady move of users to a Windows based environment has continued. Until recently two word processors have been recommended and supported but it has recently become clear that Word for Windows was the clear favourite despite PC Support's preference for Ami Pro. It has therefore been decided to standardise on Word for Windows and to promote this as a laboratory standard. As well as being the system in use in the Directorate a variant of Word for Windows is also used on Macintoshes so this should bring about the possibility of some unification to RAL word processing.
The CSSG NFS server (Bolivia) is now operational and it is being used as an extension of the PC's normal filestore as even with 200Mbyte it is very easy to fill your disc. An automated system based on the RAL virtual tape protocol is being used to backup Bolivia. An anonymous FTP server has also been installed and this will take over from the IBM as the main distribution means for PC software. The availability of both an NFS and Novell file server have allowed tests where both are in use. Although they will coexist, if you do things in the wrong order then you can give major problems on the network and some investigations into this area have been done.
The move to subnetting at RAL and the consequent change to the DNS address has highlighted the need to keep in contact with users of networked PCs. Although PC Support has records of what it has installed some installation work has been subcontracted to competent users and a large amount of public domain software has probably been installed by users. Documentation on forthcoming networking changes and how to set the main configuration variables for a variety of TCP/IP packages has been placed on Bolivia and notices of the DNS change posted on CMS and PROFS. Users of the PC/TCP are being notified directly (PC-NFS users are not affected) by paper mail and all TCP/IP users are being invited to supply their e-mail address so that future information can be passed electronically. The lack of a mechanism to put e-mail into the internal mail when a user does not have an e-mail address is a deficiency which will save us a lot of work when (if?) it is remedied.
The documentation for PC/TCP giving details specific to RAL users is now complete apart for minor changes to give details of the new documentation on Bolivia. Equivalent documentation for PC-NFS users should be ready by Christmas. The total of PC/TCP Plus installations is now 55, five more than we have licences for, and a further 17 are in the offing. A further 50 licences is now being purchased as a matter or urgency. PC-NFS installations are not yet done by PC Support but this will change when documentation becomes available.
The announcement of Windows for Workgroups by Microsoft and its likely bundling onto PCs introduces yet another PC networking flavour which, since it is effectively free, users will want to use. It thus seems likely this will create yet another isolated group of networkers alongside those using Novell and NFS. Two copies are on order for evaluation.
Viruses continued to appear with four incidents being recorded. One quite serious one affected 8 machines and another involved PC Support in cleaning over 50 discs. One virus arrived on a portable that had been to Turkey and was so new that only the most recent virus detection programs could spot it. This more than justified the cost of frequent updates to our antivirus programs and underlines the need to keep up to date in this area.
An unexpected VAT refund and the general availability of funds has prompted PC Support to reequip itself with the latest Viglens and to send the displaced machines home with staff thus allowing the home use of Windows. Given the present shortage of staff and the pressure of work this is a major way in which significant expertise can be acquired in this area. It is becoming obvious that PC Support is not able to tackle the very complex problems that can result from using Windows because it has neither the expertise nor the time. We are now starting to pay for the significant under staffing of late in that we are unable to give the more advanced users the quality of support they need.
The CD-ROM player has proved to be a reasonable buy and has been timely in that software is now gradually appearing on this medium. Corel Draw can now be run from it thus avoiding filling the PC's disc and a recent copy of AutoCAD had a CD-ROM of examples. Windows NT is only shipped on this medium which is some indication of present trends.
Version 1.2 of PKTMUX, the program that allows multiple IP protocol stacks to run over one ethernet card, has now been released. Planned developments to include more efficient 802.3 support awaits user feedback. MOS v2.3a, which includes support for PC/TCP and the latest version of PC-NFS, has also been released.
None
Recruitment of at least one new member of staff as quickly as possible.
The significant event of this quarter was the resignation of Kevin Hoadley which represents a major loss of networking expertise that will be very hard to replace. As it took nearly two years to find a replacement for Kevin's predecessor we could be in for a very lean time.
The IPTAG project to evaluate CLNS has now come to halt and is waiting for the JNT to decide whether to fund the equipment needed to make further progress. K Hoadley has continued as secretary of the IPTAG and SuperJIPS groups. He has also been involved in writing the operational requirement and evaluating the tender documents for the RAL site FDDI service.
A decision to move to subnetting was made by LMC and this has significant ramifications. The most immediate is the need to renumber both RAL DNS' as they are in the Informatics address space and this must be vacated as it is illegal under subnetting. This requires every machine using DNS for its name mapping to be reconfigured as each DNS is changed. Once subnetting is implemented there is also the need to change the network mask on every machine using IP. Proxy ARP will allow this to be done in a staged manner but it will still require significant effort, especially in the area of PCs, and the potential for mistakes and consequent network disruption is high.
NFS for CSSG staff is now operational on Bolivia and is proving very successful. However although most operations are automated it is yet another machine that has to be looked after. Its success may lead to other groups using this method for sharing data amongst PCs. The move to their own server was only made by CSSG after the IBM NFS system was found wanting and there appeared to be no other machine able to offer such a service. Perhaps the provision centrally of an NFS service for either CCD or possibly RAL, should be considered.
The Sun SPARCserver II (Letterbox) has arrived and is being set up to run PP. It is also acting as another secondary DNS which should improve resilience since Hardy proved unreliable when its services were in demand. The need for a further Sun has been identified since there is no machine available to do development work.
Testing of PP has continued and it seems to be performing to specification. There appears to be a problem with mail headers not being in the right order but this does not seem to affect mail delivery. Tests on the POP3 mail system have been less satisfactory with sufficient problems to suggest it would be worthwhile to evaluate another implementation of the server software on Directory.
World Wide Web (WWW) has been installed on Directory and can be accessed either by purpose designed programs or a normal Telnet session. It provides a hypertext style mechanism for indexing data either held locally or on other WWW systems. The possibility of developing this as a depository for networking documentation is being evaluated and some has been mounted. This can be accessed by logging into DIRECTORY as user "netinfo".
The MACREG saga continues. Plans to merge it into the DNS have been shelved since a better option seemed to be to install it in the WWW system. The move from a bridged to a routed network may reduce the relevance of MACREG since the MAC address will often not be available to identify the source of a packet. Given the reluctance of Village Managers to put their data into MACREG it might be opportune to review the situation and consider whether the time is ripe to install a network management system of some form.
LPD development is now complete for the moment. An LPR has been written to run over PC-NFS as the latter does provide this application. PC-NFS now provides most of the functionality of PC/TCP and is free. Unfortunately it seems to be less reliable than PC/TCP so the latter is preferred where the user is not short of funds.
K. Hoadley departs.
Recruitment of a replacement for K. Hoadley ASAP.
10.1.1 NEWS and Information servers
Paul Cahill installed INN (InterNet News) so that News could be transferred from Informatics with a view to taking over the current news service fed from Warwick University. It was realised, however, that AIX 3.1.5 would not adequately support this service and it was decided to await the installation of AIX 3.2 before proceeding further.
ASIS, the software repository from CERN, was installed on UNIXFE. This system provides program binaries of a large number of general applications which might be shared by a number of local systems. It is intended to reduce software maintenance and rationalise disc space usage, supporting RS/6000, HP, Sun and DEC platforms. Currently, ASIS is mainly used by two HP machines in PPD, which NFS mount specific parts of ASIS. However once AIX 3.2 is installed on UNIXFE, the system will be made more widely available.
Andrew Sansum has installed a WWW client on the IBM 3090 to allow access to the Web. He is also investigating the possibility of installing a WWW server with the aim of providing assorted information on RAL services to WWW users at external sites. XNEWS has also been installed on the IBM and provides access to news from a number of external sites.
10.1.2 Site mail services
Smail 3.1.28 was installed on Freya (a SUN) by Paul Cahill and configured to forward all mail to and from Freya via directory, the PP gateway. This was used to assess Smail as a means of improving mail provision to and from workstations on site. Mail from UNIXFE is currently sent via HHCP. There have been several instances where mail items queued for HHCP transfer have remained stuck in queues. No apparent reason was found for the problem. This situation will continue to be monitored until the problem is resolved. UNIXFE continues to provide no SMTP support although this matter was investigated by Paul Cahill who believed that Smail might prove a suitable Mail Transfer Agent (MTA) with access to both HHCP and SMTP.
10.1.3 New workstations
Two RISC based NCD colour X terminals, an NCD19c and an NCD17cr, were purchased and installed. These machines are booted from UNIXFE using NCDware 2.4. This system also provides fonts and XDM access for the terminals. The NCD17cr was provided for Stuart Martin for mainly UNIXFE based work and the NCD19c was provided for Chris Roberts for DEC (VMS) access. Both systems have LAT and DECnet support although no DECnet node numbers have been requested. Nick Hill advised that such node numbers were currently in short supply and it was felt that the terminals could function adequately under their current configuration using IP. User Support will be quite happy to share their experiences of X terminals and demonstrate these machines to other groups which express an interest.
Purple, a disk-less SGI Indigo workstation which used NFS to mount its filing system from Magenta, has been upgraded to a stand alone machine with 500MB SCSI hard drive.
There are a number of Sun 3 workstations, currently being used within the department as X terminals. These machines boot from Xerxes, which also acts as a font server, and are supported by XDM on ARCU. It was generally felt, however, that ARCU is overloaded and, with a view to moving this service, an experiment was carried out to run XDM from UNIXFE. This has been made possible by the provision of XDM as part of the NCDware 2.4 package supplied with the NCD X terminals. The trial proved acceptable and once a suitable default session is agreed, UNIXFE will become the default XDM server for these machines.
Gareth Smith of PPD approached US and suggested that the groups should share their experiences of X terminals and perhaps set up a team to further investigate the matter.
Remote access to mainframes. Following a recent announcement that NQS could be freely distributed, researchers at Daresbury expressed a desire to submit jobs directly to the Y-MP8 using the Network Queuing System (NQS). With a view to providing this option as a future service, Roger Evans agreed that Cliff Noble at Daresbury could act as a guinea pig for the scheme. This raised the problem, however, of allocating NQS Machine IDs (MIDs) across networks so that MID conflict would not occur. A scheme was devised by Paul Cahill to allocate MID ranges for sites based on their IP address range. Once this scheme is accepted by concerned parties, NQS access for Daresbury will be installed.
VTP was installed on Freya to provide access to virtual tapes on the IBM. It was hoped that this mechanism might provide a means of backup. However, the throughput of the system proved to be very low and it was agreed to wait until IP access to the IBM was improved.
10.1.4 Mailer Support/EARN gateway
Regular support work for network users continued over the period, mainly from Rita Pocock, Ann Walter and Jonathan Wheeler. On most days about one to two hours work was required to clear mail files forwarded to userid DEADMAIL. On most days there were two or three requests for help with e-mail sent to the User Support service machine (id US); usually these are requests for help in finding e-mail addresses.
Jonathan Wheeler has also been involved with sorting out problems with sites that do not accept FTP calls. Some sites have had configuration problems with their mailer service; one node (UK.AC.LON.NHLI.PRIME) can no longer accept mail and following discussions with the postmaster there, mail for this site is regularly removed from our queues.
* Investigate and evaluate X terminals in collaboration with PPD and publish our findings.
* Run XDM for the remaining Sun 3 X terminals on UNIXFE rather than ARCU.
* Install backups for workstations within the group using VTP once improved IP access to the IBM is available.
* Install WWW server on IBM and UNIXFE. Make available RAL and Atlas Centre information via WWW and publicise WWW access.
* Update NQS MIDs on local machines and publicise the change in accordance with Paul Cahill's proposed global MID allocation scheme.
* Install NQS mods to facilitate password validation and tape unit allocation requests.
10.2.1 Tasks awaiting the installation of AIX 3.2.
* Install INN on UNIXFE to provide a Usenet News service.
* Install Smail on UNIXFE and direct all outgoing UNIXFE mail to the PP gateway in order to provide both Greybook and SMTP mail.
* Make ASIS available as a UNIXFE service.
The FDDI equipment has given no significant problems. See 2.1.9.
A tender for 5 FDDI routers and a concentrator was issued during November. It would have been useful to have had further discussions with possible suppliers but this was not possible if the payment deadline of this financial year was to have a good chance of being met. Five tender responses have now been received and evaluation is about to start. The costs cover a wide range from well below the £100K target to over twice the target.
An installation working party has met. The two critical items are the delivery of the routers and the installation of the fibre optic cables. Deliveries of both items are targeted for mid January. Otherwise there are no obstacles to installation.
It has been decided to subnet the current class B IP network. The alternative would have been to apply for a set of class C networks. The class B options means least disruption to the existing network. The principle change is that all Informatics machines will have to be renumbered. See section 9.1.
Four further DEC concentrator ports have been ordered for the local FDDI network. Two machines are currently connected to the concentrator but a number of other machines are expected to be connected soon.
A decision on the FDDI router and concentrator will have been made, the equipment should be delivered, lines should be installed and equipment partly installed. Testing should be well advanced.
The Informatics machines should have been renumbered in preparation for the FDDI service.
Further VAX FDDI connections will be made.
The DECnis 600 router will be connected as an experiment.
As for last quarter, the connection of the RS6000 is still to be tested and brought into use.
It is expected that Rutherford will be about to be connected to Super-JANET. This is expected to be accomplished with a CISCO router provided through the JNT. The connection will either be to the FDDI backbone or the local FDDI network depending on further study. It is expected that all initial Super-JANET connections will use IP.
An EISA PC FDDI board from 3COM is expected courtesy of 3COM for evaluation. A suitable PC has been ordered.
The need for FDDI monitoring equipment is being studied but so far the need is unclear and there is only a rough idea of the equipment available. It does look expensive.
The UNIX front end to the Cray, UNIXFE, is now the responsibility of Systems Group.
No problems with the Maths Group RS6000.
Following the experiment with an FEI-3 connection to the Cray XMP, ARCU has been moved back to the Graphics Area and the original kernel restored. The system has proved more reliable subsequently.
No problems with the X-kernel server, XERXES, or the associated 3/50's.
No problems with the X.500 pilot and PP mail server machine, DIRECTORY.
A new machine has been purchased to act as a site mail hub running the PP software. This machine is a SPARCServer 2 with 32MB memory and 424MB of disk and is called LETTERBOX. The details of this project are covered in the TFOM minutes.
The attempt to install OSF/1 on a DECStation 5000/240 failed. This release does not have the appropriate device drivers for the 240 although it will run on a 5000/200. DEC have assured us that a release will be available shortly though I suspect that OSF/1 for the Alpha based workstations will take priority. In the meantime I shall try to install OSF/1 on the DEC 4000 AXP.
The amount of support work continues to increase and this has had an adverse effect on project work. This situation will be exacerbated by the departure of Kevin Hoadley to ULCC and Nick Gould to CERFACS in the new year.
I attended the HEPIX meeting at CERN, details of this meeting may be found in the Standard UNIX Environment committee (SUE) minutes.
Installation of OSF/1 on DEC 4000 AXP.
Upgrade DECStation 5000/240 to Ultrix 4.3.
Installation of FDDI board into ARCU.
"The Future Of Mail" working party has a number of meetings. This has resulted in a number of working papers which now define in some detail a future strategy. A final report has yet to be published.
In broad terms the aim is to move the site mail service to the "PP" machine. This includes the move of the name look up service and the EARN gateway.
Some practical work is already is progress which is reported in other sections. Quite a bit of effort has been redirected from the maintenance of the IBM services to the development of PP.
Progress expected in December 1992 January/February 1993
Further studies are needed to determine how to replace the IBM LISTSERV. The UNIX LISTSERVs are not thought to be satisfactory. Other options are to wait for EARN to develop a possible product or to adopt MAILBASE which does not interact with the world wide LISTSERV network.
Further development and studies will result in "PP" being in a position to take over many IBM services.