Predictive load balancing algorithm for cloud computing

Abstract

Cloud computing is used to provide memory and processing unit over the internet on pay-per-use basis. During the initial stage, the cloud computing environment faced complex problems due to multiplexing number of requests with available resources. The current state of the art is researchers have numerous solutions to resolve issues in handling CPU and RAM but, to the best of our knowledge; they have not considered an important factor, bandwidth. Also, many load balancing algorithms dont consider burst-conditions, which are serious issues for Cloud computing service provisioners since increase in burst time overhead results poor service time and response time. In this paper, the solution for these aforementioned problems, by using an algorithm called Predictive Load Balancing Algorithm (PLBA) is presented. PLBA is as flexible as possible, it can be used in burst and non-burst periods with maintaining service quality and consumes less amount of energy. PLBA predicts the resources before-hand and avoids overheads in non-burst periods. Also, it uses Right Scale Algorithm (RSA) to consolidate Virtual Machines (VM) into fewer physical machines. In bursts it applies round robin, to cope with overheads. We show that, PLBA can adapt in burst and non-burst conditions, with optimum use of physical machines and maintains the Quality of Service (QoS) without violating the Service Level Agreements (SLA). The system predicts the response time and service time have improved in both burst and non-burst periods with better quality of service.