The comparison is made on the basis of a model of a central facility supplying a service to a large department with 30 users concentrated one location (within 0.5 mile radius). The equipment required for a 4.8 Kbits/sec service is shown in Figure 1.
For 30 users, 3 channels are required with 6 terminals plus a hard copy device. The terminals are shown connected to a concentrator/multiplexor which would in practise be a small computer, as buffering is required to run the storage-tube displays at 4.8 Kbits/sec. A single tariff T leased line with 9.6 Kbits/sec modems connects the concentrator to the centre. A front-end-processor (FEP) is shown de-multiplexing the line into three ports on the central computer. The capital cost is £9.2K for modems + £8.6K for the concentrator = £17.8K. As the cost of the FEP is highly dependent on the particular central processor chosen, the costs associated with this part of the system are best included as part of the central computer hardware costs. Thus for terminal connection at 4.8 Kbits/sec there is a capital cost per channel of £17.8K / 3 = £5.9K.
The annual recurrent cost is made up of the line rental plus maintenance charges. Taking an average line length of 100 miles gives a rental of about £2K per annum and with maintenance at 8% of capital adding £1.4K per annum, this gives a total recurrent cost of £1.1K per annum per channel. For an equivalent service with terminal connection at 1.2 Kbits/sec multiplexing modems running at 4.8 Kbits/sec can be used. The capital cost to connect three terminals at one location at 1.2 Kbits/ sec is about £6K, giving a capital cost per channel of £2K. The line costs remain unaltered giving an annual recurrent cost of £0.8K per channel. Table 1 summarises the cost for the two speeds.
| Capital Cost per channel £K |
Recurrent annual cost per channel £K |
|
|---|---|---|
| 1.2 Kbits/sec | 2.0 | 0.8 |
| 4.8 Kbits/sec | 5.9 | 1.1 |
The Technical Group were of the opinion that the advantages of terminal connection at 4.8 Kbits/sec justified the extra capital cost and costs appropriate to this type of service have been assumed in the report.
Costs for a central service have been calculated as £5.9K per channel capital and £1.1K annual recurrent.
To calculate the cost for a multi-user mini, it is assumed that the terminals do not require modem connection and that a similar configuration to that shown in Figure 1 is used with the mini in place of the concentrator/multiplexor and taking on the multiplexor/ concentrating function. In this case it is also assumed that a 4.8 Kbits/ sec transmission speed to the centre is adequate for the reduced traffic. The capital cost for communication is £5K for modems giving, for a 6 channel mini-based interactive system, a capital cost of £O.8K per channel. The recurrent cost for line rental and maintenance is £O.4K per annum per channel.
The communications costs for single user minis are calculated on the basis that a concentrator/multiplexor would be required. Taking a system with 6 single-user minis sharing a concentrator connected to the centre at 4.8 K bits/sec, the capital cost is £5K for modems + £8.6K for the concentrator = £13.6K, giving a capital cost of £2.3K per channel. The recurrent cost for line rental and maintenance is £O.5K per annum per channel.
These costs are summarized in Table 2.
| Capital Cost per channel £K |
Recurrent annual cost per channel £K |
|
|---|---|---|
| Central system | 5.9 | 1.1 |
| Multi-user system | 0.8 | 0.4 |
| Single-user system | 2.3 | 0.5 |
It should be emphasised that these costs are only illustrative as many different configurations of terminal connections are possible.
The report envisages that in the latter part of the five-year period under consideration network communications will become available. In this section the costs involved in transforming the configurations considered above in Section 1.2 to use network connections are considered. The new configurations are shown in Figure 2.
It will be seen that, with the exception of the central computer, the hardware involved is little changed (the only change being in connection between the multiplexor or the multi-user mini and the communications line). Network connection should be borne in mind in the initial purchase of this equipment and in this case little further cost should be incurred in converting to network connection.
Conversion of the central computers will involve capital expenditure. Substantial front-end-processors will be required to allow these systems to be connected to the network without making major changes to the operating system. In addition new modems must be purchased for 48 Kbits/sec connection to the network. The capital costs for each central system will be
| Front-end-processor | £20K |
| 2 × 48 Kbits/sec modems | £10K |
For a large central service more than one 48 Kbits/sec connection between the FEP and the network might be necessary.
Line costs are now made up of two parts:
Clearly the length of leased lines will decrease, but there will be tariffs associated with the use of the network. At present there is no information on network tariff structure and in this report it is assumed that there will be no abrupt change in recurrent transmission costs as network connections are introduced.
It is possible, and very desirable, that a more integrated approach can be made at the university end and that the SRC engineering requirements can be combined with requirements of other disciplines and with those of the Computer Board. This could result in a university having an internal network terminating in perhaps a 48 Kbits/sec transmission to the network. The Technical Group recommends that the SRC, Computer Board and DES should collectively discuss networks and raise their requirements with the Post Office.
Should such an integrated approach occur there will still be need for modems for connections within the university system.
The following is a summary of the main advantages of networks for engineering research.
A network allows a piece of software (applications package, special-purpose language, etc) installed on one machine to be accessed from all parts of the network. Given the large cost of mounting and supporting software, a network clearly allows a much larger range of packages/languages to be put at the disposal of all users than would otherwise be economically possible.
With network communications terminal equipment no longer has to be dedicated for the use of one particular computer. This increases the variety of equipment (graphics devices, printers etc) available for use with any particular machine and increases the utilisation of the equipment.
Research workers with experience of using existing networks (for example ARPANET) place great emphasis on the advantages a network offers in allowing cooperation between widely dispersed research groups. Networks definitely reduce the time for new ideas to disseminate through a research community.
The user is able to choose the machine best suited for any particular job (on grounds of processor power required, memory required, software availability etc). There is also the corollary that computer managements have greater flexibility in discouraging unsuitable use of their machines.
This possibility has great value in other disciplines but is thought to have only limited value in engineering research.
In the USA this appears to be a powerful argument with cost reductions of 75% or more forecast (and to a limited extent achieved). However, it is not clear that this will be the case in the UK, although there will certainly be cost savings for the infrequent user of communications facilities.
