Evaluation and Improvement of Load Balancing Using Proposed Cuckoo Search in CloudSim

Abstract

Cloud computing offers many services by permitting end users the usage of infrastructure (like networks, servers, storage), platform (like operating systems & middleware services), & softwares (like online gaming) delivered by cloud providers (like Amazon, Google) at short cost. Furthermore, Cloud Computing permits its users to operate resources in on-demand manner. It is based on the pay as you go pricing model that causes cloud users to pay corresponding to user’s requirement only. But, demand of Cloud computing is increasing. So, Cloud Service Providers want to provide high Quality of Services (QoS) to cloud users but at same time they want to minimize cost. There are many challenges to achieve this aim. Optimum load balancing or load scheduling is one of technique which helps to achieve the Quality of Service requirements of cloud users which majorly includes Cloudlet Response Time, Cloudlet Makespan improvement. This paper consists of evaluation and improvement corresponding to these two parameters by utilizing and comparing Randomized scheduling, Round Robin scheduling, Shortest Job First Scheduling, Genetic Algorithm and last but most important Cuckoo Search With Levy Flight Algorithm scheduling results. Out of these five algorithms, genetic algorithm and cuckoo search are meta-heuristic where cuckoo search is considered to give global search due to levy flight and considered as most applicable in all application areas because of single parameter Pa. On the basis of comparisons a new Cuckoo Search has been proposed, implemented and showed better results in comparison to all above approaches. These experimental results have been attained with the help of CloudSim 3.0.3 toolkit.