An adaptive load management service in federated cloud platforms

Recently, cloud platforms have become a promising distributed paradigm for providing flexible computing capability to various high-end applications. Unfortunately, traditionally load management service can not meet the requirements of cloud platforms due to its elastic resource provision characteristic. In this paper, we present a novel adaptive load management service which applies the fuzzy inference model to implement the load-dispatching rules. In this way, we can quantitatively describe the highly nonlinear workload patterns through a small number of simple rules and then make robust load-dispatching decisions. Extensive experiments are conducted in a campus-oriented federated cloud platform to investigate the effectiveness and performance of the proposed load management service, and the results indicate that it can precisely capture the runtime characteristics of workloads and significantly improve the performance of load management service in large-scale cloud environments.