It's been the big trend in enterprise telecoms for the past five years - run your data and voice traffic on the same IP network. Yet many of the companies that have adopted voice-over-IP have decided to run separate networks for other data. Does this mean the business case for convergence doesn't stack up?
At its simplest, the convergence argument states that running one network is cheaper than running two. There's one set of routers, one set of cables and only one set of technologies and standards for your staff to understand. You only need to buy and manage one wide area network.
However, an IP phone network has other benefits. All your intra-company calls across it come at no extra cost. IP phones have all sorts of enhanced features, like follow-me roaming, presence applications and instant messenger compatibility.
Most interestingly for many businesses, though, is that IP phone systems make it easy to provision and move users. On a normal system, to move a number from one desk to another requires the services of an engineer, which can cost up to £100 a time.
With an IP phone, a user can just log in to a phone wherever in the office he or she feels like sitting down and the number will instantly be assigned to that desk.
For many companies, this portfolio of benefits is persuasive. But integrating the telephone and data networks is a step too far.
The argument simply boils down to cost. They will already have made a considerable investment into their existing phone and data networks and can't justify ripping one out to replace it all with something else.
Tim Bishop, director of marketing strategy at Siemens Communications, says: "The larger corporate organisations recognise that they have got quite a substantial infrastructure and to provide a reliable level of service they would have to spend a considerable amount of money. When they do the sums it is cheaper to do a low capacity high availability network for voice than to upgrade the whole thing at once."
Jason Angelus, product manager at Cisco, says: "Customers who want to go with a cost effective switching network, if they have already paid for all the cabling, sometimes it makes sense to do it that way."
Voice doesn't necessarily require a lot of bandwidth, it just requires that bandwidth to be guaranteed. And in some cases, the cheapest way to do that is to run it over a separate network.
Siemens's Bishop says: "10 Mbps ethernet is dirt cheap and adequate for a voice only network."
This is an approach which particularly works for companies mainly interested in IP to cut down on the costs of provisioning employees, he says.
"We have some clients in professional services who are often moving teams and phones and the costs of moving and changing are extraordinary for them," says Bishop.
Many companies considering the idea of converging all data onto one network can't quite believe it will be reliable enough.
Rightly or wrongly, data networks aren't seen as being the most reliable things in the world. And if something should happen to the data network, workers want to be able to know they can pick up the phone and call the help desk. When they pick up the phone and there's no dial tone, it becomes a very serious problem.
So there's a perception, justified or not, that phones which are running on a potentially dodgy data network won't have that cast-iron, unshakable, 'fine-nines' reliability that we're used to from conventional telephone networks.
Of course, the reliability of any given network doesn't just depend on the technology. It's as much if not more a function of network design. The more redundancy and fail-safe features you build in, the more reliable it will be.
But that comes at a price and obviously the more you spend on a network, the less attractive the cost-benefit analysis becomes - as saving cash was the reason for converging your networks it in the first place.
To an extent, it's more a question of psychology than science. Putting all your eggs in one basket just feels unwise, particularly if that basket is as shaky as the average enterprise data network.
Ian Cox, principal analyst at research firm TelecomView, says: "If you talk to an IT manager about a network with five-nines' reliability, they would look at you strangely."
Siemens' Bishop says: "We have clients whose network design allows for two separate networks. For some of them there are security concerns. They want to protect their voice network from everything that is going on over their data network."
Telecoms equipment vendors argue such an approach is wrong. Paul Rowe, senior manager at Nortel Enterprise EMEA, says: "What's the value of two separate networks? Generally our customers are looking to upgrade their data networks anyway, so it's pragmatic to run them both over the same LAN and get the benefits of IP within a converged infrastructure."
Ultimately, the convergence trend is not a revolution but a gradual migration. And like wildebeest migrating across the Serengeti plains, they may find themselves in some surprising places along the way.
Neal Tilley, solutions marketing manager at Alcatel, says: "Our customers are increasingly saying, 'I need to have the most seamless non-invasive migration possible'. They are looking to make the change to IP but because they already have networks in place they can't afford it."
"The process of converging has to take into account the current situation," says TelecomView's Cox.
"And where you have separate admin and budgets and networks, the process of converging those support budgets and networks, the process of converging will take several stages. And it may be that companies will converge at the desktop first and move onto the other stages later."