In Hindsight

When preparing for the installation, we considered (with great nervousness) what the future requirements of the Department might be. The University already has expertise in fibre optic cables - they are used for the campus high speed network, which currently uses FDDI. However, we could see little likelihood of fibre optic cabling being a general requirement within the Department. That is not to say that it may not be wanted in some specific situations, but it seemed uneconomic to try to get fibre to every "workplace".

BT offered, and we seriously considered, an approach in which empty tubes would be deployed at the same time as the twisted pair wiring. When the need for a fibre connection became apparent, the tubes would be coupled along the intended route, and fibres blown from one end to the other. BT staged a very impressive demonstration of this technique, and we recommended the University to consider it, but it was dropped on grounds of limited budget.

We think (and hope) that this will not be a serious loss: by the time that any general requirement for higher speeds might materialise, we anticipate that equipment will be available for carrying 100Mb/s data rates over our twisted pair wiring.

Certainly, at some point the Departmental Ethernet will become saturated: there is no evidence of that yet, but the acquisition of workstations, and micros with Ethernet interfaces, is proceeding at a much greater rate than we ever could have predicted when planning the wiring only a few years ago. I am thankful that we convinced the University to do the structured wiring at all, instead of installing point-to-point wiring for each specific data requirement as they had originally proposed(!)

Compared with the previous technology, i.e thin-wire co-axial cables, the structured wiring has been almost totally trouble-free. Some readers may be able to make a comparison with CERN or DESY, where an extensive infrastructure of thin co-axial Ethernet connections had already been deployed, and is now being replaced in favour of the twisted-pair scheme, for all the same reasons: reliability, versatility, resilience etc. We also make a comparison with our student labs, where, for reasons that made sense at the time, the PCs were connected together by a daisy-chain of thin co-axial Ethernet. Even with the original complement of PCs, this was far from trouble-free, and, as the number of PCs increased, the problems multiplied. Mostly when there is a bad contact, or a connector gets inadvertently disconnected, this does not only disable a single station, but makes the entire daisy-chain fail, and, now that access to the LAN is an integral part of the PC environment, rather than an incidental, this has made the cluster unacceptably unreliable. There are, admittedly, more-reliable ways of connecting up thin-wire cables, but they are significantly more expensive, and although the use of twisted pair hubs was considered uneconomic at a time when they were costing £100 or more per port, one can now get unmanaged twisted-pair hubs for a per-port cost of £15 or less, and the use of thin-wire co-axial daisy chains must now be regarded as a false economy.

We believe that in our particular situation, when the Departmental Ethernet reaches saturation we will have no great difficulty in curing it. The traffic within the different groups of the Department is much higher than the traffic between them, so by a combination of bridging and routing we can probably keep abreast of developments. The structured wiring gives us considerable freedom in assigning suitable "bedfellows", whereas daisy-chains of co-axial cable (the older technology) snaking through the premises would not. But that is probably common knowledge to you all by now. (I dare to remark here that we have benefitted greatly by coming late!) One group already set up a filtering bridge between themselves and the rest of the Department (although that was more because they already had one, brought from their previous premises, than because there was a proven need to use it). At some point we would expect to add a second Ethernet interface routed to the campus FDDI network. I really dare not predict what happens after that. Maybe we will all be ready to use ATM by then!