Swedish Institute of Computer Science  
Intelligent Systems Laboratory
Project Electronic Markets (Emarkets)
 A SITI Internet3 project
Objectives
  1. To use methods of modern finance in automatic markets for allocation of bandwidth, to avoid congestion and allow for differentiated QoS.
  2. To develop methods for fast and efficient pricing of complex network services.
Results
  1. A simulation platform
  2. A price-formation algorithm in automatic markets that minimizes communication overhead
  3. "A Price Dynamics in Bandwidth Markets for Point-to-point Connections", SICS-T2001-21
Participants
Abstract
We develop theory and methods for modeling and controlling computer networks, consisting of entities that trade network capacity between one another. Our approach allows both network providers and end-users to efficiently tailor network services to their needs. The theory can be used to compute fair prices for very complex communication needs, such as virtual channels, batch transfers with deadlines, quality-of-service, and more.

End users and network providers can express their needs in bids. Resources will be more efficiently used, and needs more closely satisfied, if markets allow for bids closely corresponding to true needs. In resource markets, e.g. in bandwidth markets, needs may be very complex, i.e. of alternative virtual paths. Automatic trading allows for liquid trading in more markets, and on much faster time-scales, and therefore entails a better global need satisfaction.

There is in general a trade-off between how complex bids a market can handle, and how quickly the market can handle them. We have created a very fast network resource market model, in which simple resources are traded in individual markets, operated by market makers. For achieving speed it is essential to only accept bids at market on this lowest level. Complex resource bundles are treated as derivative contracts on the underlying bandwidth, and are priced as such using standard mathematical finance methods. This structure allows us to price almost arbitrary complex network resources in a decentralized, efficient and scalable fashion.

We have developed a simulation platform for analysing aggregate behavior of networks used by bandwidth-trading agents.