Adaptive PSO Algorithm for Resource Allocation to Large Scale Data in Heterogeneous Cloud Computing Environment

Abstract

The resource management plays a key role in cloud computing environment to minimize the load balancing issue. In the era of artificial intelligence, a large number of workflows are generated to process on cloud servers. Efficient resource allocation in cloud computing users’ incoming tasks is one of the major issues. Efficient resource management not only improves overall system performance but reduces computing cost, degree of imbalance etc. Therefore in this paper, we have proposed an efficient resource allocation algorithm namely APSO. The proposed APSO algorithm is designed with the help of the SMIW inertia weight i.e. exponential decay factor and adaptive c1 and c2 coefficients. The adaptive social and cognitive learning factors improve the exploration and exploitations whereas the inertia weight makes balance between exploration and exploitations to find best solutions. To prove the capability of the proposed APSO algorithm, the experiments have been conducted 20 times. Based on these experiments, the proposed APSO reduced 14% makespan in case of CyberShake workflow, 14% in case of Montage workflow, 10% in case of Inspiral workflow, and 11% in case of Epigenomics over its competitor IPSO. In case of cost, the proposed APSO reduced 49% cost in case of CyberShake workflow, 53% in case of Montage workflow, 29% in case of Inspiral workflow, and 10% in case of Epigenomics over its competitor IPSO. In case of degree of imbalance (DI), the proposed APSO reduced 27% DI in case of CyberShake workflow, 31% in case of Montage workflow, 12% in case of Inspiral workflow, and comparable in case of Epigenomics over its competitor IPSO. Also, the proposed APSO improved higher throughput in all cases.