Performance Analysis of Cloud Based Application Using Neural Bio Inspired Efficient Resource Provisioning Techniques

Abstract

There have been major changes in the technology and computing paradigm since last couple of decades. Now, the focus has been shifted to the service oriented based computing paradigm and Cloud Computing is one of the example based on such computational paradigm. Resource allocation and task scheduling are prime concerns in the cloud computing environment. The efficient utilization of resources includes task scheduling at the virtual machine level in the cloud computing environment. This work achieves optimal resource utilization and minimum task completion time with the resources optimal global operational cost. newlineThe thesis work presents a study on job scheduling algorithms for Virtual Machines in a cloud based environment. The presented resource provisioning or task scheduling policies were also included on BB BC Cost, power efficient Genetic Algorithm, Artificial Neural Networks GA ANN, and fault aware based neural bio inspired approaches that addresses the tasks scheduling on virtual machine in depth. newlineThe results exhibit that the proposed cost aware BB BC approach improves the average finish time by 19.18 percent. The performance metric average resource cost improves by 30.46 percent while comparing against the cost aware Genetic approach. The proposed power aware GA ANN model improves power efficiency by 13 percent, scheduling time by 77.14 percent, total execution time by 36 percent, and fault aware GA ANN model improves fault rate by 82.63 percent. Thus, the proposed optimized BB BC neural bio inspired techniques improved performance criterion significantly in comparison to the existing static, dynamic, and meta-heuristic provisioning techniques. In the future, the proposed methods could be implemented with the Kubernetes platform, which may improve the real cloud scenarios performance. The performance criterion will include reliability, cost, and time. The presented methodologies will also be tested for host level scheduling. newline