الفهرس | Only 14 pages are availabe for public view |
Abstract Cloud computing (CC) was emerged as a computing paradigm, which aims to offer reliable, customized and Quality of Service (QoS) guaranteed dynamic computing environments for end-users. The data centers in CC are growing tremendously in order to meet the rising demands such as rapid computational response, massive storage, etc. Virtual Machine (VM) is extensively adopted as an enhancer of CC, which provides several benefits such as performance isolat io n, security, flexibility and ease of management in the user customized platform. These benefits impact the resource utilization, performance and power consumption of the data centers and could be able to minimize the maintenance cost. In the task allocation process, Software Defined Networking (SDN) is used to present poweraware dynamic allocators for virtual machine. The drawbacks of the prevailing algorithms could be overcome with the optimal dynamic features of Earliest Deadline First (EDF) algorithm. In this framework, our study introduces 10 VM with various allocation methods and compares them with a baseline approach containing first available fit. These allocators vary in accordance with allocation policies, strategies and other network resources. Task scheduling is a method used for allocating the tasks to server on the basis of workload capacity. In general, the tasks are allocated to the corresponding server for minimizing time delay and traffic. Particle Swarm Optimization (PSO) is one of the best algorithms utilized for task scheduling in cloud platform with low computational cost. In this study, a Hybrid Swarm Optimization (HSO) has been proposed, which is the integration of Salp Swarm Optimization (SSO) and Particle Swarm and Optimization for resolving the prevailing complexities. The predominant goal of the proposed framework is task scheduling of the resources available for reducing the computational cost and execution time. Multilayer Logistic Regression (MLR) is a technique for the detection of overloaded VM, thereby task could be scheduled to the virtual machine in accordance with the workload capacity. The proposed HSO with multilayer regression was simulated in cloud sim toolkit. The output of the proposed framework depicts the effectiveness of the proposed system in termsof cost, makespan, and execution time. When compared to the prevailing systems such as Improved Efficiency Evolution (IDEA), Genetic algorithm (GA), and PSO, the performance of the proposed system is shown to be more efficient in task scheduling and virtual machine allocation. With the proposed framework, a novel idea for overcoming the issues for task scheduling and job allocation for virtual machines in CC has been accomplished with low percentage of cost and low time duration. |