Distributed Algorithms for Providing Fairness in Heterogeneous Computer Systems

Distributed computing systems often consist of heterogeneous computing resources managed by different administrators. Due to the distributed nature, effective management of the resources may become difficult and the performance of the system may be affected. Also, the users of distributed systems may be self-interested whose goal would be to maximize their own utility. Such selfinterested agents may adversely affect the performance of the system. Here, we present performance optimization algorithms whose objective is to provide fairness to all the users of a distributed system involving selfish users i.e. all the users will experience approximately the same expected response time for the execution of their jobs or will have to pay approximately the same expected price for the execution of their jobs. Distributed heterogeneous systems with various system and node models are considered. Experimental results with various system configurations are presented comparing the performance of the presented algorithms with other existing schemes.