The Research Of Intelligent Algorithm For Task Scheduling Based On Workflow System In Cloud Computing Environment

As one of market-oriented distributed systems, cloud computing has become popular in recent years. In cloud workflows, scheduling is an important and challenging issue, especially at the task level. There are different kinds of user’s requirements should be considered to obtain the suitable resource with pay-as-you-go in cloud environment. The efficient scheduling strategy can allocate the tasks to suitable computing resources. The cost of the workflow is the key issue cared by the users. Also the providers focus on the efficiency and economy of the scheduling.In cloud computing environment, a suitable scheduling algorithm, the fundamental of scheduling optimization, can contribute to find the best scheduling. The advanced approaches and research achievements is the basis of this article. In cloud workflow system, scheduling optimization is an important issue, especially when considering about the market-oriented. The majority of optimization is about two key factors:time and cost. As scheduling is a well-known NP problem, some heuristic algorithms such as ACO (Ant Colony Optimization) and PSO (Particle Swarm Optimization) have been proposed to optimize the time and cost. As we know that heuristic algorithm is the summary and imitation of the laws of nature. The characteristics of cloud environment have to be considered when the heuristic algorithm used. There are many articles focus on the optimization of scheduling algorithm. However these articles have some weakness in optimization. This article analysis the weakness and propose new solutions for scheduling.In cloud computing environment, the scheduling execute kinds of tasks with VM (Virtual Machine). These computing resource execute different tasks in same time. The VM visualized by computing resource also have the ability to execute tasks in parallel. On this basis, this article analysis the hierarchical structure of cloud workflow scheduling, and aims to optimize the time and cost of scheduling with intelligence algorithm. In time optimization, for current scheduling strategies, the tasks execute sequentially in VMs, and each VM is shared by multiple users in public cloud workflow environment. The algorithm combined with the time-sharing VM would efficiently reduce the makespan of scheduling. In cost optimization, PSO have been proposed to optimize the cost. However they have the weakness of premature convergence in optimization and therefore cannot reduce the cost effectively. The cost of scheduling will reduce observably when solve the problem.Considering the existing problems of scheduling, the innovations of this article are as follows:(1)In time optimization, we propose the ACO-parallel algorithm for tasks scheduling with considering the time-sharing characteristic, which can minimize the makespan of task set with cost constraints. Then, we build a new makespan model for scheduling based on the time-sharing characteristic of VM. In experimental simulation, our scheduling can converge faster and achieve smaller makespan. Furthermore, with the increase of capacity of VM/number of VMs/available budget, our scheduling can always achieve a near-best makespan. It is effective to gain smaller makespan with parallel heuristics than without.(2)In cost optimization, CPSO (Chaotic Particle Swarm Optimization) algorithm with chaotic sequence and adaptive inertia weight factor are applied to solve the premature convergence. Chaotic sequence with high randomness improves diversity of solutions, and its regularity makes sure good global convergence.In experimental simulation, our scheduling algorithm can overcome existing premature convergence problem and achieve smaller cost. And then, with the increase of available deadline, our scheduling can always achieve a best cost.To enhance the optimization of current scheduling, this article aims to optimize the time and cost of scheduling with considering the market oriented characteristic. In these two optimization, the scheduling of tasks is more reasonable and the efficiency of tasks execution have been enhanced. The cost of users are reduced and the benefit of providers are increased. This article is very important both in science and economy.