Research And Implementation Of Task Scheduling Algorithm In Cloud Environment

In recent years, with the rapid development of cloud computing, more and more institutions, researchers and users are studying it becouse of its huge computation ability and convenient service mode. Cloud computing is a distributed system with massive computing power which consist of shared resources and heterogeneous resources. Cloud computing resources have characteristics such as widely distributed, self-management, heterogeneous and dynamic load change etc. Becouse of those characteristics, task scheduling in cloud computing faces more complex problems than which in traditional distributed environment.Task scheduling is mainly to assign tasks to the appropriate resource nodes to meet the needs of users on the execution time and cost. Aiming at the problems existing in the execution time of task scheduling and workflow task processing problems, we carry out the following tasks:First, in order to solve the problem of minimizing the tasks scheduling finish time, after reading a lot of papers, we proposed fuzzy clustering of resources according to the resource characteristics, then the resources are classified, in that way, we can solve the problem of large resource search time.Then, we proposed budget and deadline constraints heterogeneous earliest finish time algorithm to meet the needs of the user’s execution time and cost. Experiments show that the algorithm and strategy we proposed achieve better results.Second, according to researchers using PSO algorithm in cloud computing environment to excute task scheduling, only working well in independent tasks, we optimized the PSO algorithm. On one hand, we introduced workflow process model to consolidate the intensive tasks and reduce the particles dimentions; on the other hand we optimize the definition of particle initialization and adaptive function, to ensure that the search space and convergence speed. The experiments show that the enhanced PSO algorithm has better convergence speed and better performance in the execution of workflow tasks.Finally, this thesis design and implement the basic task scheduler on unified cloud management platform. And then, we embeded the task scheduling algorithm we proposed in it. We described task queue management module, task scheduling module and real-time monitoring module in detail.The experimental results show that the proposed resource scheduling algorithm is effective and feasible.