Simulation of a market model for distributed control

Decentralized control in the allocation of computing resources to tasks in a multiprocessing environment is designed using a general concept of a market. Two different market models are developed: auction and barter. In both cases, the tasks compete among themselves for computing resources rather than having them assigned by a host. Definite relationships between market strategy and system performance were found. Moreover, the results varied with machine configuration and appear applicable to systems of distributed control.