Brampton Factor: Network convergence is no simple task

As businesses rush to roll out converged networks, few are taking the time to consider whether it's really the right move for them. Martin Brampton considers the subtler elements of convergence.

Talk of network convergence is pretty odd when so much is said about IT needing to be business-oriented. Naturally, using the most cost effective technology is a business issue but it is hard to see how deployment of any particular technology can ever be a business issue. Sometimes convergence provides a good solution, sometimes it doesn't. Often other issues are more important.

Another common tendency also works against us. That is our propensity for pushing any new facility to the point where it breaks down. Adding video into the converged solution is a case in point, both for domestic and corporate networks.

In the domestic context, despite higher ADSL speeds, it is important to remember that contention means each user is only actually entitled to 40Kbps. It is also doubtful whether the internet could cope with substantial video traffic without major upgrading. The same kind of stress can arise in corporate networks when senior executives demand online access to training material and such like.

How badly do we need convergence? The answer is surely we don't need it very badly. I have a home telephone, a mobile and a Skype phone. Yes, it would be more convenient if I had a single handset that transparently used all of them - especially if we could do away with telephone numbers, Skype fashion, thus rendering our already damaged directory enquiries service obsolete.

But do I have a problem figuring out which phone to use? No. It never causes me a moment's concern. And given actual charging structures it is sometimes cheaper to use a mobile than a fixed line.

These counter-intuitive economics illustrate how, in either the corporate or domestic environment, it is important to remember that charges for communication services frequently do not reflect costs - at least not in a simple way. In practice charges give a highly distorted picture of costs, a situation that leaves providers with plenty of scope to alter the tariff in ways that are liable to alter the cost-benefit calculations at short notice.

It is essential to remember that VoIP services such as Skype calls are not free because bandwidth is never free. It may be that its marginal cost is sometimes zero because bandwidth is bought in packages but that is simply a feature of the present tariffs.

Is there a skills shortage?

Tell us what you think of the IT workforce in silicon.com's 2007 Skills Survey.

What Skype calls do provide is a voice service over bandwidth that is much cheaper than bandwidth provided by a voice carrier such as BT. But this arises simply because the cost of handling calls is now very low for the carriers. That is why providers are so keen to sign people up to inclusive packages. As pressure from the likes of Skype steadily increases, so fixed-line call charges will steadily fall. It pays to be cautious about spending money to exploit price differences of this kind because the difference will inevitably erode over time, maybe quite quickly.

Indeed commercial negotiation may be a much more cost-effective approach than technical deployment. Carriers are naturally resistant to erosion of call prices but they know it is inevitable and are still keen to obtain marginal revenue. It may be easier to get good prices by using an intermediary who buys in bulk and resells to a sizeable number of corporate clients.

Another consideration is that the majority of large networks are poorly understood. Traffic has typically been added piecemeal over a long period and much data communication does not require particularly high standards of service.

This is especially true of service-oriented architectures (SOA). Ambitious developments in this area have led to a situation where the precise communications flows are not well known. As this kind of architecture is often built to be tolerant of high latency and lost packets, nobody is aware of issues until the network is subjected to new stresses. Services such as VoIP, Citrix-style thin clients or video are not at all tolerant of low quality networks.

Understanding the network also involves questioning what is actually carried over it. Assuming a justification has been made for carrying voice or video across a data network, this type of traffic needs to have priority over low-grade traffic such as email and general web browsing. It may well make financial sense to have different routes for different kinds of traffic, since building high performance networks involves significantly increased costs.

There is, therefore, a need to optimise the mix of services that are provided so as to match cost to real importance. Thinking this way, it is evident that convergence might be the exact opposite of what is most needed.

Questions about converging voice and data in a corporate network lead to consideration of management of service levels. Very high standards are expected for telephony, which has typically operated to extremely high levels of availability, sometimes in the absence of mains power.

Even with careful design, some form of quality of service management is usually needed to ensure voice is handled to an acceptable standard, maybe in the presence of adverse conditions from the data that shares a converged network. In anything but a green field site, this is likely to generate substantial costs, as network hardware of various ages is commonly encountered, offering different levels of capability. Rebuilding the entire network is an expense that is not always justified.

Bearing in mind the tricky issues that are not automatically solved simply by installing a converged network, I remain unimpressed by organisations that publicise their networks as a badge of quality. It is more impressive if decisions are made on the basis of a sound assessment of economic impact and to a high technical standard. Surely that is what IT being business-aware ought to be about?