|Homepage / Publications & Opinion / Silicon.com
Peter Cochrane's Uncommon Sense: Circuit or packet, clean or dirty?
...networks and the internet, that is. Did you know the dilemma facing telcos worldwide?
The move from circuit- to packet-switched IP networks leaves communications companies in a strange position. Can they upgrade and trust voice over IP? Can they afford not to, asks Peter Cochrane?
I was recently asked if the telephone network will be wiped out by the internet. Will voice over Internet Protocol (VoIP) win the battle? Will video conferencing over Internet Protocol (VCOIP) go mass market? My take on both? Eventually but not overnight.
The predominant mode for all real time human telecommunication is circuit switching, with a dedicated connection between two terminals for their sole use throughout the session. When you pick up a telephone and dial a number, a circuit is established between you and whoever you call and it lasts for the duration of your conversation. The same is true for video conferencing and all real-time broadcasting.
For non-real time applications such as email and delayed voice and video we can use packet switching where the digital information is broken up into small packets and directed one at a time, across the planet, using any available routing at the time of despatch. Instead of a known and dedicated connection, each packet may arrive by a vastly different routing.
This subtle difference between circuit and packet switching was established by technologies of two different eras. When the telephone system was first implemented the electronics to do anything sophisticated are unavailable and even the thermionic tube - the valve - had not been invented. Everything was electro-mechanical.
It was the need for computer communication that created the packet switched philosophy to obviate the need for 100 per cent dedicated circuits. At that time such connections were very expensive and computers inherently need to communicate with multiple destinations at the same time for seconds and not point-to-point for minutes.
Today we communicate and the majority of us have absolutely no idea how the circuits are configured or indeed how they operate. And why should we? But just stop for a moment and ask yourself how you would engineer a system that would allow any two people out of a 1,000,000,000 population, distributed over 200 countries, to connect and speak at will? This is not a trivial problem.
The actual solution is currently in a critical cost and performance battle between two distinct philosophies - circuit and packet switching. In most towns you will find a telephone exchange that occupies several floors of a large building. In that same town you will find an ISP, providing a packet switching capability in a single room, or indeed a single equipment cabinet. As the bits per second passing through these facilities are of the same order, then the promise is that huge (and very expensive) circuit switches (occupying buildings) could be replaced by something that would easily fit into your garage.
If you ever have the chance to experience a VoIP telephone call or VCoIP take advantage of that chance and watch and listen for the artifacts. It is impossible to control the internet in the same way we control the telephone network and so congestion, collisions and traffic jams cause delays, bunching of packets in an unpredictable manner.
A huge amount of effort has been expended trying to cure this problem. The most successful attempts see the prescriptive routing and priority of packets so the end-to-end circuit, although based on packets, rapidly resembles its circuit-switched predecessor. For many people the quality of the speech and the vision provided by IP technology is adequate but for professional use it will never make the grade and there is now an invisible and unvoiced movement to create a second internet.
It has been widely proven that VOIP and VCOIP can be engineered over private intranets where the bandwidth and the collisions (traffic jams) can be controlled. So many companies are having their circuit-switched private networks based on PBXes replaced by IP-based networks (intranets) on which computers and people can communicate. One of the secrets for success is an overall reduction in the number of hops from one point to another and the strict control of the traffic to give priority to all real time applications.
It is clear that the telephone network of the planet cannot be replaced by the uncontrolled and chaotic internet. But the joining together of purpose-engineered intranets could edge us slowly towards a world where no one produces a circuit switch machine ever again. But the reality is that no one has a model that allows us to engineer this today.
So industry faces an interesting predicament where no one is willing to manufacture the big telephone switches anymore and no one quite knows how to replace them with the packet-switched equivalent. In the meantime, the old switches are getting older and not being replaced. The clock is ticking for an industry born of the 19th century.
I think we are about to see the emergence of a new and clean internet, one that is controlled and not subject to hacker attack or subversive use, but maintained by companies and organisations for the express purpose of driving down the cost of telephone and video communications and providing a greater utility than is available hitherto.
The IP takeover is well underway and in the US companies increasingly offer me a 'dirty internet' connection as a visitor while employees enjoy a 'clean internet' facility that has replaced the 'dirty internet' and PBX. For IP, it is not a question of if, but when.
This article was dictated whilst driving to Heathrow on the M25, typed by my PA a week later, edited on AA512 flying San Diego to Chicago, and polished a few weeks later on my laptop over a hot coffee and a pile of flu pills, to be despatched to silicon.com over a newly installed broadband link from my home.