The Research Of Task-level Scheduling Algorithm Of Workflow System In Cloud Computing Environment

With the in-depth development and research of cloud computing, there are a growing number of scientific workflow, business workflow and collaborative applications which have been developed in a cloud computing environment. They usually are powerful and require a large number of resources. Meanwhile in the cloud the process of application services is becoming more complex. In addition, they are subject to the constraints of the tree factors (cost, time and resource). By Visualization Model, Cloud workflow can flexibly and quickly build complex processes, then run and manage the cloud computing applications according to the created process, so that the cloud application services can be automatically and efficiently executed. Compared with other traditional computing environment, users access to computing and storage resources according to their own need and pay the money in accordance with the sources they have used. Because of the unique nature of cloud computing, the traditional workflow technologies can not be a good solution to the cloud workflow management problems.Task scheduling and resource allocation are two important technologies in cloud computing. Cloud workflow scheduling refers to that each task of the cloud workflow instance is assigned to the appropriate resources and managed running. It can directly affect whether the cloud workflow instance can be successfully and efficiently executed or not. Unlike in other traditional computing environment, cloud workflow scheduling not only considers whether the best resources chosen meets the pre-defined task scheduling constraints (typically consider running time and running costs), but also has to pay attention to causal link between the various tasks, as well as to coordinate the implementation of each task to get the best results of the implementation. Cloud workflow scheduling usually is NP-complete problems.In this thesis, we focus on Task-level scheduling in Cloud Workflow and analysis the features of virtual machines which is from the underlying resource by virtualization. Considering all kinds of QoS constraints of workflow tasks, we propose ACS scheduling algorithm based on virtual machine sharing features. The algorithm takes into account the overall tasks cost constraints, and optimizes execution performance, as well as sets the allowed maximum number of parallel tasks on each virtual machine which is from the underlying resource by virtualization, considering the performance of it. Because Task-level scheduling in Cloud Workflow deals with integrated workflow instance, QoS constraints of each task are more complicated. We set a variety of heuristic information for the constraints. By the simulation experiment, the proposed algorithm, compared to the other algorithm,shows the better performance in the situation where there are many parallel tasks, and can make good use of virtual resource and optimize the Task-to-VM assignment in the cloud data centers.