Evolutionary Approaches for Load Balancing in Cloud Computing

Abstract

Recently, cloud computing has emerged as one of the prominent technology among other newlinecompetitive technologies such as grid computing and wireless sensor network. It has newlineextended its empire in IT industries for its abundant opportunities. It provides cheaper newlineservices and resources irrespective of the machine, location, and time. Due to the flexible newlineservices, available, and flexibility, this paradigm is adopted by researchers, IT developers, newlineand common users. newlineHowever, the development of a different mechanism to dynamically balance the newlineload across the globe is the common problem in this system. There exist many associated newlineissues related to load balancing. In this thesis, few evolutionary approaches have been newlineaddressed for load balancing. Each contribution is dedicated for load balancing in a cloud newlineplatform. The performance of the algorithms is evaluated using a simulated environment newlinecalled cloud analyst. newlineIn the first contribution, a Multi Particle Swarm Optimization (MPSO) based met newlineheuristics load balancing technique has been proposed. The Particle Swarm Optimization newline(PSO) has been used two times for better exploration of search space. The random initial newlinepopulation is generated and given as input to the first stage of PSO. The output generated newlinefrom the first stage of the PSO along with some solution generated through the newlinedeterministic method and random mutation is considered as the input for the next stage of newlinePSO. The comparisons are made with First Come First Serve (FCFS), Round Robin (RR), newlineMin Min, PSO, and Genetic Algorithm (GA) with different measures like makespan and newlineaverage response time by varying the cloudlet, virtual machines (VMs). The proposed newlineapproach achieves better load balancing in large scale cloud computing environment. newlineIn the second contribution, JAYA algorithm is used for load balancing to overcome newlinethe constraints of algorithm specific parameters and the common parameters used in most newlineof the evolutionary algorithms. Further, the proper tuning of the algorithm-specific newlineparameters is a very crucial factor which affects the performance of the above mentioned newlinealgorithms. Jaya Algorithm used very less number of controlling parameters for newlinecomputation. The performance comparisons are made with GA and PSO with different newlineviii newlinemeasures like average response time, data center service request time, data transfer cost newlineand virtual machine cost etc. JAYA algorithm outperforms other approaches. newlineSimilarly, in the third contribution, a population-based algorithm called Forest newlineOptimization Algorithm (FOA) is used for load balancing in the cloud computing system. newlineThe comparisons are made with PSO, GA, and JAYA algorithm in different measures like newlineaverage response time and total execution time by varying the number of tasks and the newlinenumber of virtual machines. It is obtained the FOA performs better when the load ratio is newlinehigh in comparison to other competitive approaches newline