A structured argument

Increasingly, ten Gigabit/s Ethernet over twisted-pair copper cabling is taking a centre stage role. Customers are beginning to request this form of structured cabling, which extends the capabilities far in excess of any previous copper cabling formats.

However, the standards surrounding ten Gb/s Ethernet have yet to be ratified (see box, ‘Setting the standards') and it is true to say that the market is still at an early stage. For today's contractors and installers, the situation is becoming confusing: do they stick with standard Category 5e and 6 horizontal cabling systems, with a fibre backbone; do they start to offer ten Gb/s Ethernet; or make the switch to fibre-to-the-desk?

History in the making

There has been an IEEE standard for ten Gb/s Ethernet since 2003, specifically over optical fibre. It has been successful in the wide area network environment, but has not penetrated much into the enterprise network. This is largely because the bulk of higher data rate applications use fibre channel (FC) protocol, rather than Ethernet.

Also, ten Gb/s Ethernet over standard 62·5/125 fibre will only perform over cable runs of 35 m; although there are plans to extend this capability.

So, attention turned to delivering ten Gb/s Ethernet over twisted-pair copper cabling (10GBASE-T). The single biggest advantage is the reduction in cost. The expectation is that the active equipment for the copper links will be about half the price of that for fibre links when the standard is launched.

The main focus of 10GBASE-T is storage and server applications. It is also potentially suitable for backbone network applications.

Another standard - 10GBASE-CX4 - came out in 2004. This is designed to be a lower-cost alternative to optical fibre for rack-to-rack links, switch clustering and some data centre applications. It can only support up to 15 m lengths of cable, so application is limited.

Aside from the fact that the standards are not finalised, there are other issues to consider. With 10GBASE-T, copper will be running closer to its physical limit than ever before, which is likely to make a system more difficult to install and maintain. The bend radius will be highly challenged. The cable is also likely to be heavier and more expensive than standard copper cabling. Until the standards are issued and components can be fully qualified, interoperability between manufacturers cannot be guaranteed.

However, the biggest issue is probably alien crosstalk (AXT), namely signal coupling from one channel to another. In 10GBASE-T, the signal will be carried equally over the copper pairs, with each pair supporting transmission in both directions.

Since the loss of the cable rises steeply with frequency, the signal to noise ratio (SNR) must be kept as low as possible. Although the standards bodies are looking at how they can address AXT, there are still unknown factors about this complex topic. And despite extensive modelling, as yet there is no finalised method of measurement.

Noise reduction

Cables are traditionally bundled closely together, meaning that the noise from one can couple into others. This has been a problem in the past, but it has been manageable. However, due to the sensitivity of 10GBASE-T it can no longer be ignored.

One answer is to space cables out, rather than bundling them in neat parallel lines; another is to separate channels altogether. This does create some potential issues: the installer and user must understand the system and maintain it in the ‘as installed' state. Also, the cabling will take up far more space.

These complicated mitigation techniques can be avoided by using fully shielded cables, jacks and patchleads that are compliant with the new standard. This should allow up to 100 m channels to be installed without the use of mitigation techniques for AXT.

Another alternative is to make the move to fibre optic cabling. It may be relatively expensive, but it is light, small in size and immune to EMI and RFI. Further, prices could drop in the future.

Looking at the bigger picture, it is important to ask whether customers really need ten Gb/s Ethernet yet. After all, most users have only recently starting using 10-100 Mb/s and 1 Gb/s represents a small percentage of installations today.

Yet it would be dangerous to take a Luddite's view of progress in this area. As has been seen in the past few years, data bandwidth demand has grown far in excess of most people's expectations and early adopters already want to be ready for 10 Gb/s. It is important for contractors and installers to be aware of all the facts surrounding 10GBASE-T before they make any decisions.