An agent-based model for price dynamics in a GRID market

Uli Harder

Department of Computing, Imperial College London Huxley Building, 180 Queens Gate, London SW7 2AZ, UK

This work presents an example of agent-based models for demand in economic and social systems with a market environment. In particular we are investigating how a market system can be used for GRID computing. GRID computing environments are already in use in scientific areas like particle theory. Recently this idea has also received increased interest from industry, especially in the banking and games area (IBM). There is a consensus that soon there will be GRID companies providing computer power in a generic way, like power and telephone companies provide their services nowadays. Therefore it is paramount to investigate how such a market would price its products. Also, it is important to have a tool to investigate different scenarios with a wide range of market mechanism to explore what the best arrangement is. One problem any analytic or simulation approach faces is that it is difficult to get realistic user behaviour (i.e. demand) for the a model. However, in the context of car traffic Nagel and Schreckenberger proposed a very successful model. Also, for stock markets models have been suggested by Bak et al. In the context of the GRID the main focus has so far been on the implementation of charging schemes, though there have been attempts to model price developments in a GRID market. We are suggesting an agent-based approach to the problem. In our model we have a number of GRID companies selling their computing through brokers to end-users. The end-users have different strategies on the selection of services provided by the brokers. Similarly brokers have different ways to achieve their aim of making profit.