Network Intelligent or Stupid
by Thiyagarajan Maruthavanan (Rajan)
Rajiv mentions in his recent post about an interesting discussion that surfaced in the mobile monday meet about design of network architecture. Where should the intelligence lie in a network? It is a quite an important discussion and I will add my two bits to the meme here.
As Rajiv pointed out that those who come from the telco world revere that intelligence should lie in the network. I think that is the case because the people who from that side of the camp have not been exposed to any alternate network design structure so as to evaluate against anything else. Secondly we humans are in general resistant to change of any kind and also why would someone want to jeopardize the huge pot of gold one owns. But then those who come from the PC (software) world would think who would be such a bozo to put intelligence(in other words optimize the core) into the network.
From technical PoV Is it better to design a network where the core of the network is intelligent and the edge of it very dumb ?
End to End Principle in system design is a seminal technical paper written way back in the 80’s by Reed, Schultz & Clark which argues why it is better to design a system in with intelligence kept to a minimum at the core. The argument in the paper is that
“ that functions placed at low levels of a system may be redundant or of little value when compared with the cost of providing them at that low level”
To understand this in an intuitive manner reflect on this, today’s optimization for someone is another person’s headache tomorrow.(Software engineers can relate to this very well as they face this situation a lot in their daily work). By putting an intelligence (in other words optimizing for something) we reduce our option value for a later application tomorrow. For example the PSTN network that was built was optimized for voice totally and it can’t be used for any other purpose. To support a different application a new network has to be installed (which involves massive cost).This is also explained in a very clear way in another seminal piece of work by David Isenberg. He wrote about it as an article in 97 titled as “The rise of stupid network”.
Thus technically speaking it is better to design a network which is stupid network at the core & the intelligence should be left to the edge. If you look at networks that are gaining prominence & acceptance by users then they are the stupid networks ( internet, Japan’s i-mode). If it were not for a stupid network then during the recent ban on blogspot sites we would have found ourselves very helpless. Even when ISP’s blocked blocked the site we could very easily circumvent that. A stupid network puts maximum power into the hands of a user that is why millions love it are adopting it with great fervor. It makes living on the edge possible & so much fun 🙂
The best network thus is a stupid network as far s user is concerned but will this network will help a telco milk money out of users is totally different question.