Research On Load Balancing And QoS Oriented Multi-objective Cooperative Task Scheduling In Cloud Environment

Task scheduling is one of the most critical issues on cloud platform. The number of users is huge and data volume is tremendous. Requests of resources sharing and reuse become more and more imperative. Cloud Servers consistently make response to users, during which involves users’ job scheduling and management. How to efficiently dispatch resources, schedule tasks and return a better task sequence, maintain the whole system load balance so as to avoid hotspot and finally improve resource utilization, has been a research topic in cloud environment. It’s a fact that resources stay heterogeneous and change dynamically in cloud environment. Besides, it’s a key issue that how to efficiently and reasonably dispatch users’ tasks to different resources according to the Quality of Service (QoS) requirements of both cloud computing centers and users. Long time no response will surely lead to expensive rent cost and dissatisfaction.To solve above problems, considering the new characteristics of cloud computing and original adaptive genetic algorithm(AGA), different optimal objectives oriented JLGA(Job Spanning Time and Load Balancing Genetic Algorithm) and QoS-GA(QoS Oriented Genetic Algorithm) adaptive double fitness genetic algorithm are proposed. The algorithm designs multi-objective cooperative task scheduling schema. JLGA not only works out a task sequence with shorter job makespan and average job computing time, but also satisfies inter-nodes load balancing. Considering QoS requirements in cloud environment, QoS-GA is proposed, which adopts double fitness adaptive genetic algorithm-job spanning time and rent cost. At the same time, this paper adopts greedy algorithm to initialize the population, and weights multi-fitness functions. Finally, we compare the performance of JLGA with AGA, and QoS-GA with AGA respectively. Receptively simulation analysis of JLGA, AGA and QoS-GA, AGA shows that JLGA outperforms AGA in term of job response time and load balancing, and QoS-GA performs better in both time cost, total rent cost, and is more suitable for users. JLGA and QoS-GA are better adapted to the cloud computing environment.