After years of being told that everything will eventually go IP and run on the same network, are the limitations of that approach starting to show?

At issue is the fact that networks are increasingly becoming application-orientated, and different apps have different needs. For example, while many apps are highly tolerant of latency, others are not. And some, such as voice over IP, require predictability within the switch.

The result is that some network architectures - and some switch designs - could be more appropriate to certain apps than others, and vice versa.

Network manufacturers are realising that, according to Mark Slater, who has just moved from Extreme Network's UK office to take over as the company's new corporate technical director in California.

"A lot of infrastructure is still built for data quality, not triple-play," he says. "For example, it doesn't scale to too many MAC addresses, or it doesn't do quality of service.

"All networks are not equal. For data networking they're pretty much equal, but for anything above that they're not. People buy networks on specifications and don't test enough, and tenders rarely cater for the complexity of the intended use."

He adds that as more and more services move onto IP, the desire for service level agreements and interdepartmental chargeback means that network management must differentiate between the needs of different customers - in short, it has to get a lot more granular.

"It's the next level of triple-play [data, voice and video], the ability to differentiate per customer and treat each independently," Slater says.

And of course, while those capabilities may be new to the enterprise, there is one area of business where they are well established - the service provider.

So, as new apps converge onto the IP network, network managers need to do two things: first, start thinking like a service provider, and learn how service providers work.

Second, ask more questions, and try to test the hardware and the apps together yourself. If you can't, find someone else who's running them already and go ask them what went right and what went wrong.