Monday, January 24, 2005

Can I trust my telephony to IP?

(This article appeared in Networked Comms Insight during 2003)

Up until a decade ago, I used to work on the dark side, deriving most of my income from Telecoms manufacturers. I was involved with many new developments, generally at field trial (beta) stage, where the work tends to be with Customers who are early adopters, who actually want the solution being offered.

What it is easy to overlook working for a large Corporate in a globetrotting, ground breaking environment, is that sometimes developments are solutions looking for a problem. This is where the magical world of marketing takes over, in persuading Customers they may have a problem that their solution is optimal towards solving.

Packet switched telephony is just such a solution. I was reliably informed by clever people that circuit switched technology was a dinosaur back in 1988, but here we are 15 years later and whilst it is now starting to look very old and tired, it still hasn’t become extinct yet. Why is this?

Well, speaking as a Telecoms Manager, 50% of the sector seems to have a vested interest in my being sceptical. IP from the new boys leaves something to be desired, they hint. It might be quite reliable now, but a lot of those features that you take for granted just aren’t there. Never mind the quality, feel the width. Of course, our solution gives you the best of both worlds, everything you want, in an IP environment, fully integrated with your legacy stuff.

Hang on, though, I think. You guys have been charging an arm and a leg for functionality that has paid for its development many times over. If I want to add clever stuff in my old and new bits I still have to pay through the nose for it. If I don’t you deem it to be legacy and I have to pay catch-up to do the slightest thing a few years down the line.

But hang on again, all these licensing and legacy things seem to be a very similar scenario to the way the IT Software industry has gone anyway, so I am going to have exactly the same issues with the new players on the block. Lets take a closer look at their story.

The idea of convergence, it seems, is to bring voice, data and applications together in one cohesive whole, a true Information & Telecoms Technology (ICT), or perhaps better expressed as INS, or Information and Network Services. One network saves on costs and is a great enabler for multi-channel, unified functionality. Bob Emmerson, the well-known industry watcher has coined an even more interesting term that embraces the applications, connectivity and process in a holistic manner. He calls it UCI, which stands for Unified Communications and Information. So, yes, there is a definite problem that IP can be an enabler for solving; our communications channels are islands of functionality isolated from our processes and each other.

Well, that sounds great. However, not all of us are blue-chips with huge budgets. So what do we need to do to start trusting our voice to IP?

Our biggest hurdle is the Network.

Firstly, we have to tackle the problems of packet loss, delay, jitter etc. Voice doesn’t need massive bandwidth but it is hugely intolerant to any mucking about. This means the network has to be fully switched from edge to edge where you want to talk, with 100 Meg to the desktop and Gigabit core. It needs to support packet prioritisation, probably at both layer 2 locally and layer 3 when routed.

Prioritisation by itself doesn’t help in overload situations. A bit too much voice will dramatically impact on the non-voice throughput; way too much voice could cause bottlenecks and quality deterioration. Whilst bandwidth planning isn’t constrained by real channels any more, there is still the need to plan and engineer the network.

Resilience is crucial to continue to deliver high availability that phone users have always expected. It is a good idea to provide UPS further out towards the network edges so that phone conversations are immune to power cuts. Hang on though, many IP phones have power bricks into 13A sockets. Fortunately, 802.11af power over Ethernet has now been standardised. Always ensure there is adequate provision of Ethernet power, whether via in-line adapters or switches that include it. Given time, in-line power will become standard rather than an up-lift and one less variant to manage.

Once the network is voice ready, we then have to take stock of our legacy voice systems. They are also a big hurdle, because their legacy gets in the way of a UCI approach. Some of the vendors are starting to show what we might call green shoots of UCI, as there are some interesting partial solutions out there. The Cisco’s and 3Coms of the world also have their own UCI offerings.

I think that having an IP phone on every desk is a bit of a red herring in the majority of Companies. Yes, making the configuration follow the phone means that the instrument can go into the packing crate along with the files when someone moves and the functionality follows them. However, we have to get away from the mentality of the number belonging to the phone. When we move, we want to leave the thing on the desk, log in and then the associated phone becomes ours, for the duration, or by default if we choose it to be (semi) permanent.

I want us to get away from having a separate phone and PC, but I’m not talking about kludges like soft phones, USB interfaces or sound cards. My ideal workstation looks like a PC with a handset, headset or bluetooth associated with it. However, here is the crunch. The phone is actually built into the keyboard and is the network connection, powered via Ethernet. A small display & message waiting light above the numeric keypad is the only immediately obvious difference, along with the cradle, if fitted. When we arrive in the morning, we have a fully working traditional phone, however, once the PC has booted up and we have logged in that is when the phone really performs, being fully integrated with all our other communication channels and applications.

What about all of the users who just need POTS? Initially, put IP to analogue gateways in the closets and patch them as distributed analogue devices locally. There comes a point when IP-POTS will be cost effective enough to spend £50 on a simple IP phone and not have to manage them as a special case. In the IP world, vanilla telephony (PIPS) can include a display, message waiting light and half a dozen feature buttons, as the marginal cost between production runs of nasty phones compared to basic ones will indeed be marginal.

Having moved on to futures, let us return to the present. There are two further obstacles to achieving voice UCI. Security is an obvious one & there are (possibly) urban myths about denial of service attacks, hacking & viruses taking down IP phone systems. There are also stories of full voice IP systems running physically separated from the main data network, not just Vlan which is the virtual equivalent. Security has to be ubiquitous and multi-tiered, a point not to be laboured here.

Once we have ourselves a secure, resilient, reliable, robust IP network, our final hurdle is to manage it. The move, change & feature bit is probably a lot easier than what the voice people have been used to. However, what is the really critical bit is network management, both from a real-time and historic perspective. With the success of Cisco AVVID, there seems to have been a clutch of enterprising small companies who have popped up to provide useful tools outside of the scope (& huge cost) of the traditional tools like HP Openview, Cisco Works, Transcend et-al. We also have to work out how to get our voice, application & data staff trained up with the right skill set to do it, or find a partner who will.

Will I trust my telephony to IP? Well, I would if I had the time, money and resources. In the meantime, I have one IP phone and am thinking about installing another one! Now if only the IP-PSTN were here today…

Bob Emmerson can be found at

No comments: