A Load Balancing Policy for Distributed Web Service

The proliferation of web services; and users appeal for high scalability, availability and reliability of web servers to provide rapid response and high throughput for the Clientsʹ requests occurring at anytime. Distributed Web Servers (DWSs) provide an effective solution for improving the quality of web services. This paper addresses un-regulated jobs/tasks migration among the servers. Considering distributed web services with several servers running, a lot of bandwidth is wasted due to unnecessary job migration. Having considered bandwidth optimization, it is important to develop a policy that will address the bandwidth consumption while loads/tasks are being transferred among the servers. The goal of this work is to regulate this movement to minimize bandwidth consumption. From literatures, little or no attention was given to this problem, making it difficult to implement some of these policies/schemes in bandwidth scarce environment. Our policy "Cooperative Adaptive Symmetrical Initiated Dynamic/Diffusion (CASID)" was developed using Java Development Environment (JADE) a middle ware service oriented environment which is agent-based. The software was used to simulate events (jobs distribution) on the servers. With no job transfer allowed when all servers are busy, any over loaded server process jobs internally to completion. We achieved this by having two different agents; static cognitive agents and dynamic cognitive agents. The results were compared with the existing schemes. CASID policy outperforms PLB scheme in terms of response time and system throughput.