Impact of Adaptive Resource Allocation Requests in Utility Cluster Computing Environments

Abstract

Maximizing resource provider profit and satisfying user requirements at the same time is a challenging problem in utility computing environments. In this paper, we introduce adaptive resource allocation requests and investigate the impact of using them in utility cluster computing environments. The Service Level Agreements established between users and resource providers rely not only on fixed values, but also on functions that associate allocation parameters. In addition, the resource provider scheduler can automatically modify the number of resources and usage time of allocation requests, as well as split them into subre- quests. Users may receive incentives for supplying flexible requests which produce more scheduling options. By using rescheduling, resource providers are able to prioritize the most profitable requests dynamically and still satisfy the requirements of the already accepted user requests. From our experimental results we observed an increase of 14% in the resource provider profit and a reduction of 20% in the average response time of user requests when compared to traditional approaches.