In this paper, we discuss two major tradeoffs, spatial and temporal tradeoffs, that appear when applying market-based computing to multimedia network applications. The spatial tradeoff exists between computation and communication cost, depending on how agents in a market are distributed over network. The temporal tradeoff appears between reactiveness and correctness of computing equilibrium, depending on how agents adapt to dynamically changing network environments.
We analyzed the above tradeoffs considering two extreme applications. One is the bandwidth allocation in FreeWalk multimedia conferencing tool, and the other is that in RSVP environment of the Internet. As the result, we clarified that (1) as for a spatial tradeoff, the distributed computation becomes profitable as the number of clients increases, and that (2) as for a temporal tradeoff, the merit to respond quickly to the environmental change supersedes the merit to improve the accuracy of the resource allocation: the former prematurely terminates the computation of market price, and the latter performs the calculation until the market perfectly clears. From these results, we concluded that (1) the centralized computation fits to FreeWalk environment where the number of clients is small, while the distributed computation fits to RSVP environment where their number is large, and that (2) it is necessary to stop the iterative calculation at some suitable point both in FreeWalk environment where the demand for QoS changes quickly, and in RSVP environment with a large number of clients where it is not practical to wait for the accurate equilibrium.