Abstract
Nowadays, cloud computing and big data are changing the enterprise. Cloud computing, as a new business computing mode, distributes computing tasks across resource pools made up of a large number of computers for large-scale calculation. In the current research on the task assignment problem of cloud computing, most scholars consider single-objective programming, for example minimizing the cost or makespan. However, factors such as load balancing of virtual machines, reliability of system, and energy consumption can influence the quality of the cloud computing service. Therefore, in order to adapt to the development of practical applications, multi-objective programming should be considered in the task scheduling problem of cloud computing. Cloud computing refers to storing and accessing applications, data or services over the internet remotely. Due to its large server size and wide user base, cloud computing is faced with two major problems: reasonable allocation of cloud computing recourses and efficient scheduling of a large number of application tasks. The task scheduling problem of cloud computing is essentially an NP-hard problem of combinatorial optimization which is hard to find the global optima within polynomial time complexity. In recent researches, heuristic algorithms (such as genetic algorithms and particle swarm optimization) are used to solve the problem. Such methods can hopefully make use of the strength of heuristic algorithms and find high-quality solutions within an acceptable running time. This paper proposes a new algorithm (Multi-Objective Simplified Swarm Optimization，MOSSO) for multi-objective problems, based on a new, convenient and efficient heuristic algorithm called Simplified Swarm Optimization (SSO) and using the procedure and idea of establishing a repository in the Multi-Objective Particle Swarm Optimization (MOPSO). In order to increase the search ability of feasible solution space in this algorithm, this paper designs dynamitic parameters to make the mutation rate large at the early stage to enhance global search ability and the mutation rate small at late stage to enhance local search ability. In order to verify the performance of the proposed algorithm, this paper uses randomly generated cloud computing data for simulation analysis and makes comparison with other algorithms. The final result proves that the method proposed in this paper can take both efficiency of algorithm and quality of solutions into account, and play the strength of Simplified Swarm Optimization to help users find the best quality solution and finishing multiple objectives such as minimizing total power and minimizing makespan at the same time.