|Homepage / Publications & Opinion / Articles, Lectures, Preprints & Reprints (post 2002)
Bottom Up Wins, Top Down Loses!
12th September 2005
For well over a decade it has been obvious that Internet Protocol (IP) and Voice Over IP (VOIP) were going to become a significant threat to existing networks. The established PSTN way of doing things changed society and served us well for over 100 years, but it is now way past it's sell by date! And the move to IP is far more significant than a mere 'circuit to packet' transformation. IP Nets involve a change in the philosophy of the way we design and operate networks from the bottom up, as opposed to top down. It is a move from the Do IT For Me (DIFM) world of hierarchy, order, control, to a world that is low, flat, chaotic and Do IT Yourself (DIY).
An IP network can replace the existing telephone network for <10% of the upfront and operating costs - in that sense it's a no brainer. The really big difference between these two modes of operation (circuit and packet) is that one automatically ensures a very definable Quality-of-Service (QoS) for each connection, whilst the other is absolutely indeterminate (in the raw state). The telephone network is centrally controlled whilst IP Nets are peripherally controlled. One has all the intelligence in the core, and the other has it all on the edge. One is the domain of the big corporation, the other, the domain of the individual. One is limited and sluggish to change, the other is fleet of foot. One is designed and planned, and the other evolves. One is governed by specific big machine design, the other by consumerism. And in short - consumerism wins!
Circuit switching in telecoms network affords a unique end-to-end connection for every call with a globally defined QoS in terms of bandwidth, signal-to-noise ratio, and the likelihood of a connection failure. In complete contrast, packet networks look much more like a game of roulette. No unique circuit is established for end-to-end connection and indeed each packet may take an absolutely different route as the traffic loading and congestion on the global network changes across the planet.
Whilst circuit switched networks were fundamentally designed to support real time communication of audio (speech) signals, packet systems were designed to support data transmission and are fundamentally unsuited for any real time activity. What we have seen in the last 10 years has been the bastardisation of both paradigms in an effort to make both do things they were never intended to do. Circuit switching is poorly suited to any data transmission needs, and so this fraternity came up with all manner of clever coding, routing, and switching techniques to extend the life of a now defunct concept. And in contrast, the packet switching community tried everything they could think of to bend a medium poorly suited to speech and video.
So where have we got to? The telephone networks of the world are moribund and destined to go. Most likely there will only be museum examples within the next 20 years! IP networks are taking over, but in a more sophisticated form than evident in the Internet version! In order to provide the QoS required for all sectors using real and non-real time services our IP networks have to become a little more ordered! Ironically, to provide good quality speech services it is necessary to nail down the packet routing for the duration of the connection. The same is true of video and conference calls. In effect the packet switched network becomes a pseudo-circuit switched network for just that period!
Personally I see the emergence of several levels of IP networking necessary to safeguard our future. First there is the Internet: Near 100% anarchy and freedom with all the dangers of Virus, Worms, Trojan Horse and Spam attacks etc, but with great utility and freedom. Second: there are the corporate Intranets: Banks and companies establishing well ordered, clean, protected and high performance networks free and isolated from all the dangers of the Internet - and without many of the freedoms. Third: I see a new breed of secure network for government and emergency use that is really bolted down and well defined, and above all, well protected and defended.
Just three years ago I was objecting to having to purchase WiFi connectivity or indeed connectivity of any kind in hotels as I travelled, but with the advance of broadband and the falling cost of provision I can now mostly find free access. But ironically, I now find myself prepared to pay for connectivity if necessary! The reason? The ensuing saving I can recoup on VOIP for fixed and mobile calls, and more especially international calls. It is not unusual for me to make a Skype connection with my son at his school, for an hour, from California at a cost of <$0.70, or indeed to conduct a conference call or radio interview across the Atlantic for much <$1. These savings alone mean that I can afford to buy WiFi connectivity and a full IP connection should the need arise.
There is however, a downer in the VOIP story. Today VOIP connections have only two modes of operation - absolutely brilliant and absolutely awful. The principle reason for this is, of course that IP networks were never designed to carry any form of real time service. To make them suitable for universal support of VOIP it really is necessary to predefine and nail down the routing of the packets in such a manner that the number of hops is minimised, and that they stay stable throughout the duration of a call. A number of companies have undertaken various engineering approaches to solve this problem and it is an ideal window of opportunity for the old Telcos.
Leading Telcos have now recognised that they have to move into the IP world really fast if they are to survive. The good news is that they own much of the infrastructure and can afford to buy the expertise required. Also, replacing their old multi-$Bn PSTN investments with multi-$M IP replacements won't be as traumatic as they fear!
When we made the transition from copper cables to optical fibre in long lines networks, the operating costs dropped to <10%. When fibre is deployed in the local loop the same thing occurs in terms of the installation, operating and support costs. The consolidation of switching and routing systems, followed by a move to IP, will see a further massive reduction in the amount of equipment and people required.
The old PSTN required switches in the centre of conurbations because the signalling and transmission limitations of copper meant that the average distance to customer was ~2.2km whilst the extremities of the network only stretched to 7-17km in Europe and North America respectively. In contrast optical fibre allows span increases to >40km. Moreover, the cost a large switch (central office) runs into many $Ms whereas a router or IP switch runs into $ks.
So the fundamental cost of networks using fibre and wireless with IP, as opposed to copper, wireless and PSTN circuit switches is vastly less. The major problem for the incumbents is writing off all of their old investments, which would normally have been written down over a period of 30 years, which have now got to go in a period of 3. The opportunity for the new players, on the other hand, is to establish networks that are much lighter weight, have greater utility and versatility and potentially far more reliable.
One of the biggest worries of this transition is that we don't quite get it right! The lessons of over 120 year of telephone company operation are in danger of being lost. New networks are being constructed without adequate battery back-up, without standby generators, without duplicated equipment and alternate routing strategies, and without adequate network performance control to realize the QoS we have come to enjoy on the past. A counter argument would say that no engineer, manager or salesman could have contemplated or guessed that the general public would have put up with such a bad services and pay so much as they have with the mobile phone. On the face of it we seem to value mobility and price over QoS!
In comparison with the fixed line telephone network, mobile phones are generally poor - well hold onto your hats because VOIP turns out to be even worse. How far can we go down that trail? I am not at all sure. But there are significant questions about 911/999 emergency services, about the ability of these new networks to support emergency communication and to create a new networks that always ensures a sufficiently good quality service with no failure characteristics. Telephone networks evolved with hieratical, very well defined and understood. IP networks are inherently evolutionary, chaotic, undefined and ill understood.
One thing that has become evident from the tragedy of 9/11 in New York and 7/7 of London, is that we do need to ensure that we have more than one network. During 9/11 several switches were taken out by the collapse of the Twin Towers, but the mobile network held up reasonably well and the Internet carried on. During 7/7 in London, it was mobile nets that froze and were in overload, whilst the PSTN continued, as did the Internet. For those who see a world of a single super highway, I would council that we need multiple networks based on IP, but built to differing standards to ensure that we don't see a common mode failure, during some tragic event that may be natural or manmade.
Mobile phones with VOIP are now on the market and I regularly use VOIP from my laptop. In addition, music, gaming, entertainment, cameras, GPS, RFID and many more services and devices are afoot. They are coming from the consumer and not from the networks. They cannot be controlled or limited, so the big question is, can IP networks rise to the challenge? I think they can? And the basis of my assertion? The Internet just gets better and better year on year with no centralized engineering control. Throwing bandwidth at the problem just seems to do the trick! And with just a little engineering to help network evolution along, I suspect the QoS change would be dramatic!
Word Count = 1,706