If anyone should be able to benefit from convergence, it's a radio station. Over the years, radio broadcasters have acquired a huge legacy of communications - satellite links for network programming, leased lines to transmitters, ISDN, telephony and data networking will all be in there somewhere.
But if you're concerned about the impact of IP on voice phone calls, imagine how much more critical it must be when your users expect CD-quality stereo sound. Yet at GWR Group, one of the UK's largest independent radio companies, a converged all-IP network is not just on the menu for the future, it has been in operation for almost a year now.
GWR operates more than 30 local FM stations around the UK plus Classic FM and various AM and DAB (digital radio) services, and is in the process of a merger with the Capital Radio group which will add yet more stations. It now has an MPLS network carries all the group's data and telephony, but more importantly it carries the audio for transmission, and has made the organisation much more flexible as a result.
"We had separate networks for telephony, data and live audio," says Vincent Bourne, GWR's infrastructure manager. "What we wanted was to be agile and able to add stations very quickly. The idea was to plug a cable in and add a few phones, and you're on the internal phone book, able to take syndicated shows, etc."
It also means that presenters can work from anywhere, he adds: "We had a night presenter who had to drive from Birmingham to Bristol every day because the satellite uplink was in Bristol. Now a DJ can work from any of our sites."
Deciding that one vendor is easier to manage, GWR chose to work with Cisco for hardware and bought a managed MPLS service from Thus. Bourne says both have been very responsive - Thus was building out its network as GWR was testing, and Cisco provided direct access to its engineers in San Jose.
He adds though that he and his team deliberately used all the leverage they had as a high profile site: "The good thing is everyone understands radio, and we used all the leverage we had - everyone involved wanted to have this story to tell.
"We spent a long time planning this. I joined from a WAN build and design background, but radio's quite immature and lacked IT investment. In the early days we threw in a standard Frame network to buy us time."
The GWR team knew that the keys to making it all work were QoS (quality of service) networking, to ensure audio quality, and multicast for syndicated shows and national content.
"It was luck that our satellite, Frame and telephony contracts all ended within months of each other - we started with those dates and worked backwards," says Aidan Hancock, GWR's network manager.
"The key enabling technology is QoS, until that worked you couldn't do this with IP and were better off buying five lines. We have five QoS classes - telephony, broadcast, critical business applications such as sales, less critical applications, and 'the rest' - email, FTP, Web.
"Telephony is interactive while broadcast is predictable and one way, the tolerances are different but the main difference is it's predictable. The other thing is the boundaries are movable and dynamic, so when one app's not using the bandwidth, others can use it. From a technical standpoint, we were gobsmacked how well QoS worked.
"Multicast was the big challenge for us, Cisco and Thus - multicast over MPLS is advanced stuff. It worked fine, until you changed something. We did find bugs in the Cisco code, but we also had access to Cisco engineers to fix them."
Resilience is vital too: "Convergence is great but you'd better make damn sure your network doesn't fail," Vince Bourne adds. "The most common issue is the average station has a two meg tail with a router from Thus, typically delivered by BT, and those sometimes go down. So we have a complementary ISDN network that dials around the cloud for backup, with lines that are tested weekly. We made our radio application bandwidth-aware as well, so over ISDN it drops to mono, leaving enough bandwidth for a few phone calls and so on."
He says though that, in many ways, the human and organisational factors were more important than the technical ones: "The technology is the easy bit - any number of telecomms vendors will sell you a QoS network. The devil is in the detail.
"From a business perspective, who should own a convergence project? Probably not the IT department - maybe the finance department? They need to understand the costs of telephony. Maybe facilities, they have comms knowledge but not IT.
"Convergence is not a pet project. You have to have a lot of visibility at board level and you need a project owner who can work across different areas and remove conflicts - a key influencer.
"Don't underestimate the vested interests either - a lot of skill-sets will become redundant. And be prepared for a lot of restructuring - we moved the voice people into the technology department."
In particular he warns against what he calls terrorist elements: people who will just dig their heels in and obstruct the project, or who expect it to fail even before it starts. And because you won't usually churn your whole infrastructure overnight, he says it's important to maintain strong relationships with suppliers and staff, especially around your legacy systems.
"We didn't lose any sleep about IP telephony because people have done it before," he adds. "It was how I funded the rest of the project. We removed all the PBXs, and all calls are now routed through Bristol, London or Reading."
"We've replaced 70 PBXs - I think every one was a different model - with three Cisco Call Managers," says Aidan Hancock. "The telephony savings represented 80 percent of the business case for the project - about £500,000.
"We have around 2000 IP phones. All the FM stations are now on the MPLS network 11 hours a day since January 04, the AM stations for 20 hours a day - the network feed has to go through the local stations too for the addition of jingles, ads and so on. It's a huge amount of comms and technology."
The next phase is to add in the Capital Radio stations, and Hancock says they expect to drive the time required down from six months to four to six weeks per station. Then they will explore bringing the transmitters into the network and complete the centralisation and distribution of content.
He says that return on investment is already coming from voice over IP and centralised phone services, such as a virtual switchboard for smaller stations. There is more to come too, from services on the phone and from handing off telephony to the PSTN as IP - he'd like to do this now, but isn't convinced the carriers are ready.
"For us it was easy," he adds. "We always wanted to broadcast using IP. It was very clear what we could do for the business using technology."