Over the last ten years, the voice telephony industry has changed beyond all recognition as the technologies of delivering voice and the technologies of 
datacomms have merged. Voice, which was once something that stood on its own, with its own rules, its own technologies, its own people and its own ways of working, came together with datacomms through VoIP (Voice over IP) and is slowly subsuming itself into the general IT industries.
By today’s standards, voice prices were high ten years ago. There were good margins to be made on switching voice minutes and better margins to be made on terminating calls in far-away places. Prices have tumbled and customer expectations of prices, even on very long distance calls are very much lower than they were. To some extent, that is driven from computer applications such as Skype, offering voice calls which are free. The consumer price expectation is therefore set to zero and every provider who competes with free has to justify in the consumer’s mind why his service should cost more than free. Clearly, that is possible, otherwise the whole voice industry would have disappeared. Consumers still pick up their landlines and their iPhones and make calls which they pay for, and perceive that they get value for the money they pay, otherwise they would not do it.
In the old technologies, a physical circuit was divided up amongst many calls on a fixed time slot or Time Division Multiplex (TDM) basis, and the information in the call was described in digital form. The internet has no physical circuits, only the movement of packets of data (TCP/IP packets). Packets may be routed anywhere, anyhow. As long as they travel between the source and the destination, the route each packet takes is of no importance. In this sense, a voice call is just the same as any other internet communications activity, and the cost of voice calls falls along with the cost of the rest of the IT and datacomms world – very fast. Voice is just another computer application, operating as a piece of software running on a standard off-the-shelf server. And the relative size of the voice industry as a part of the datacomms industry gets smaller and smaller – general IT applications demand more and more bandwidth every year, but the amount of bandwidth needed by a voice call stays constant and there are only so many voice calls consumers can make, even if the price falls to free.
But, as with most things in life, all is not as simple as it may sound. Voice is a lot more than just moving some packets around. If a general IT application wants to load a spread-sheet, for example, the source sends the packets and the receiver assembles them together to recreate the spread-sheet as soon as all the packets have arrived. If there is contention or heavy traffic or packet loss, it takes longer than on a good quality, clean link. With voice, that is nowhere near good enough. Latency (the time taken for packets to get from one end to the other) matters, but of much more importance is differential latency, ie the differences in the time delays between packets. Packets have to be re-assembled in the right time order to recreate a voice. If early packets arrive after later packets, voice quality suffers. So what starts to matter most is not the ease of buying a computer application, loading it off the CD and reading the manual, but the quality, coverage and characteristics of the carriers moving the packets in and out of the application.
All carriers are different. Some offer consistently high quality, some offer consistently high support levels, some offer very high availability levels. Some offer diverse routes to foreign destinations. Some are very cheap (but, as a general rule in life, you get what you pay for. Cheap is good, but it is unrealistic to expect good quality, good support, high reliability, diverse routes etc at the same time as cheap, no matter what the salesman says). The right carrier choice is key to voice operations.
Consider two different telephony operations. One sells number translation, one sells prepaid calling cards. The number translation operation has a very limited carrier requirement – probably two carriers or a multi-homed service for resilience and reliability, but limited geographic needs, limited latency or differential latency needs and limited bandwidth needs. It can be located in a data centre facility almost anywhere. The prepaid card operation, on the other hand, may need reliable, diverse, high quality routes to Glasgow, Buenos Aires and Lahore. The carrier requirements are very high and most unlikely to be satisfied by any one carrier. Such an operation needs to be located in an environment where there are a large number of high quality carriers on-net, with a physical, supported presence in the facility.
Which brings us round to the very beginning. The places where such concentrations of carriers are to be found tend to be in the places where the voice telephony companies were located in the old days of TDM voice. The best places to go for a wide choice of carriers to connect a voice telephony operation to the world tend to be the data centres that had them in the first place. In the UK, all of those are in either central London or docklands. If there is a general rule for locating the new off-the-shelf computer application that is voice, it is to put it where the fibre is and choose somewhere that is reliable and with people who are nice to work with.
City Lifeline
