Comms supplement: Cabling feature – Cheaper copper sails on?

In these straightened times, cheaper copper-based cabling systems such as 10GBASE-T are proving to be just as popular, if not more so, than higher functioning but more expensive fibre optic systems. Philip Hunter looks at the pros and cons of both options and asks which has the wind in its sails

Copper cabling has enjoyed more lives than the proverbial cat, and now two big factors, the credit crunch and arrival of the 10GBASE-T standard, have conspired to postpone its long expected replacement by fibre optic yet again. Recessions have always given copper a boost and this one is no exception, exposing once again the higher cost of fibre and the associated terminating equipment. Similarly, 10GBASE-T is the latest in a series of breakthroughs in physical transmission that keep copper within sufficient distance of fibre in capacity terms to make the cost and disruption of ripping it out and starting again unacceptable to many financial institutions, except perhaps wholesale banks running algorithmic trading applications, which need the speed and size of fibre. “Every time fibre has looked like surpassing copper, the copper development teams have surprised everyone, giving it new life,” agrees Alastair Waite, head of product management at ADC Krone, a vendor of fibre and copper structured cabling products. That said, fibre has been accounting for an increasing proportion of enterprise infrastructures for over a decade. This is particularly the case for some banks and trading floors where budgets have been less constrained and fibre’s advantages of greater capacity per volume of cable and other factors like immunity from electronic eavesdropping come into play. Any wholesale bank would be hard pressed to greenlight a fibre infrastructure investment today however with the downturn in full swing – you’d need a far sighted board that can see the long-term benefits.

Any progress into the copper heartland within data centres and trading floors seems to have been arrested for now by the credit crunch, and by the timely arrival of 10GBASE-T. Although 10GBASE-T does not have the range of fibre, it does maintain the existing coverage of up to 100 metres, so there is no additional motivation for migrating to fibre on this count. In any case, 10 Gigabits per second (Gbps) is overkill for most work areas, for which even one Gbps will normally provide more capacity than required. “Banks today are not foreseeing applications to the desk needing 10G to support administration or retail as 1G currently handles these,” says Waite. “The exception though is new trading floors, where 10 Gbps is being provided either via fibre or copper right to some desks, for future proofing, even though the capacity is not needed now.”

The main application though of 10 Gbps infrastructure will be in backbone networks serving the workgroups, and here the motivation is to shave as much as possible off latency to accelerate trading execution. “10GBASE-T will be adopted mostly in the data centre and in horizontal cabling, particularly to trading floors where milliseconds of advantage can translate into millions of pounds of additional profit or loss,” says Ian Wilkie, business and marketing director at structured cabling vendor, Brand Rex.

Indeed, latency has been cited as one of the major IT issues faced by many wholesale banks given this dependence on sub-millisecond trading. Paradoxically, the adoption of 10GBASE-T, designed to mitigate latency, puts the spotlight on copper, and makes it more desirable to keep the physical links as short as possible. This is because the time taken for electrical signals to traverse a link remains the same irrespective of the operating frequency. Therefore, as data transmission speeds increase, so the link length becomes increasingly significant – it’s a factor network designers should increasingly take into account.

This consideration also suggests though that 10 Gbps might after all prove to be the end of the road for copper, with fibre finally winning out at advanced future speeds of 40 Gbps and 100 Gbps, where realistically only fibre can go. “In the data centre there are already users like Google and Yahoo desperate for higher capacity interconnections,” says ADC Krone’s Waite. “The institute of electrical and electronics engineers (IEEE) is working on 40Gig and 100Gig fibre systems and my opinion is that 100Gig on fibre is the clear winner for the next phase of high-speed data centre local area networks (LANs).” NYSE Euronext, for example, is introducing a claimed 100G low latency trading and market data network with Ciena, as part of its project to launch two new data centres in London and New York in 2010.

But that is for the future. For now 10 Gbps meets the needs of most normal data centres, few of which have Google-like throughputs. Crucially, 10 gigabit architectures meet the latest specifications for storage area networks (SANs) based on the fibre channel standard, which now supports eight Gbps.

Even when 10 Gbps is not yet required it should be taken into consideration when existing cabling infrastructure is coming to the end of its life and needs replacing. The infrastructure itself will typically last two or three generations of the underlying physical transmission technology that it supports. And 10GBASE-T really requires Category 6A grade cable, which is higher specification but more expensive than the Category 6 or 5A found in most existing infrastructures. Category 6A, defined in February 2008, operates at 500 MHz, double the frequency of standard Category 6, enabling copper to support 10 Gbps at the full 100 metres.

“Network managers are now asking themselves the question – will this system ever need to go beyond one Gbps,” says Brand Rex’s Wilkie. “If the answer is no they are generally installing Cat 6 or even sometimes Cat5e. However, if the answer is yes or even maybe, the additional cost of Cat6A is so relatively small that companies of all sizes in all sectors are taking the ‘safe’ route and installing Cat6A – because the cost and disruption of having to cable the building or data centre again in less than its design life of seven to 12 years is unthinkable.” Future proofing makes sense if you think you genuinely might need the extra functionality later on.

Managing your investment
Another question companies may ask themselves when replacing existing cabling is whether to invest in Intelligent Infrastructure Management (IIM) systems. This was supposed to provide the hitherto missing link between higher level real-time network management and the underlying cabling infrastructure previously reliant on manual procedures for patching and changing. Potential benefits include lower costs and a reduced chance of downtime resulting from cabling changes, via automating the change management process and keeping track of assets.

Yet as noted by Graeme Stoker, EMEA marketing manager for Siemon, a structured cabling vendor, despite its promise IIM has not yet been widely adopted. “IIM remains a very powerful technology, but has never really gained widespread popularity, probably because many of the costs and outages that IIM can cure are relatively invisible (even though they add up to very high costs to the business),” he says.

IIM has also been a victim of the credit crunch, according to Ewan Wilson, managing director of Northampton-based HT Data, a supplier of cable management software that is part of the global HellermannTyton cabling components group. “There has been a decline in the take up of IIM systems, probably accelerated by the adverse economic climate,” says Wilson.

This may prove a false economy given that it leaves cabling the only unmanaged part of the IT firmament, and yet an important potential cause of downtime, and of latency when the network is working. “However it may just be that existing IIM products are too unwieldy and expensive,” says Wilson. “We are seeing more interest in our software-only solution, which has a lower level of investment and a ‘sufficient’ cost benefit. It has less functionality but at a much more attractive price point.”

The decline in full blown IIM also illuminates another trend that may be accelerated by the credit crunch, which is the outsourcing of data centres. This, as Wilson notes, transfers capital expenditure to operational cost (Capex to Opex), and also sidesteps cabling issues, such as whether to use copper and, if so, which grade. It is too early to say whether this outsourcing trend will turn into a stampede, but if it does that could also hasten the march to fibre as managed service providers will need more capacity than individual users.

>> other supplement features