Agent Based Load Balancing Mechanisms in Federated Cloud

Abstract

Cloud computing is one of the recent innovation in the field of information technology, which provides services to user on demand and pay per utilization. Single cloud based service model lacks in performance factors like response time, throughput and deadline missing etc., when workload becomes heavy. To overcome this limitation, federated cloud management broker architecture was proposed. Since cloud traffic is unpredictable and busty in nature, there is a possibility of large number of incoming service requests for processing. Hence the workload varies dynamically, some service providers are overloaded and others may be under loaded. In order to balance this situation, to improve the performance of federated cloud broker architecture, load balancing techniques are incorporated at the place of Broker Manager and brokers. Broker Manager (BM) plays a vital role to select appropriate best broker for computing the incoming service requests. Agent based Round Robin Load Balancing Scheduling (ARRS) is proposed at BM for assigning the service requests among the selected brokers by considering the parameters such as workload and queue size of brokers. Another one called Decentralized Agent based Load Balancing (DALB) technique is proposed at the level of brokers to balance the assigned workload in the way of migrating the requests to the under loaded brokers. The result shows that the proposed load balance based broker architecture provides better performance compared to without load balancing based architecture.