« Fun With WiFi | Main | Song for the Day »
September 12, 2003
Talk: networks and games
I attended the first technical talk I've had in a long time this past week. The presentation was conducted by Dr. Christos Papadimitriuo of UC Berkeley on the topic of computer networks, game theories, and how the two mixed can be used in conjunction to build a better network infrastructure. After looking through his web-page, I believe the discussion was a 50 minute introduction to the concepts he will be discussing in his CS294 seminar at UC Berkeley.
As a warning note there were points during the presentation that I could not hear Dr. Papadimitriuo, as such I may have missed some important details and my interpolated details may not be correct at all.
The talk begin with a basic introduction to game theory and some of the more popular concepts within, i.e. the penny game, etc. The main points I picked out of this talk revolved around the concept that, while the shortest path is always the desired route, sometimes taking the initially longer path will result in better performance, and the challenges of adding these to the current network traffic patterns. His best example used sample routes and had the attendees calculate out the optimal routes. The fastest route was easily found, not because it was obvious, but rather because we could see the future routes and predict around them.
At this point he began to discuss the challenges of adding this logic to routing, the games that need to be applied, and the theories of how/way this works. It was also at this point that an audience member challenged his theory, by stating that the theory works great but the reality is completely different (and well documented/understood). It was at this point that I could no longer hear the exchange occurring between the two, but it placed enough of a seed in my own mind to question the validity of the research. Thanks to this interaction the time frame for the lecture had to be compressed into a much more condensed version.
One of the parting questions asked was what would happen if an entity were actively trying to exploit the system. In a sense, the proposal he had given was for the induction of a benevolent monarch to control in a fair manner, and continue to push the expedient routes. The question revolved around this monarch being replaced with one who's intention was not to keep the fastest routes, but rather the most optimal? Or more importantly what would happen if one of the routes decided to act against the orders from above, how would that impact the overall network strategy? Unfortunately these questions went not really answered/explained due to time constraints.
[EDIT: typo corrections]
Posted by Dan at September 12, 2003 01:13 PM