Research On Workflow Scheduling Algorithm And Implementation Based On Workflow QoS Constraints In Cloud Environment

Cloud computing is becoming one of the hot target in research institutions at home and abroad. Cloud computing is the development of Parallel Computing, Distributed Computing and Grid Computing. It is a new business computing model which uses mature virtualization technology to package data center's resources, providing on-demand services for client on the Internet. Task scheduling and resource allocation are two key technology in the cloud computing. The commercial characteristic of cloud computing makes it focus on Quality of Service providing for user. And the use of virtualization technology makes its resource allocation and task scheduling apparently different from the previous parallel distributed computing. Cloud computing is characterized by such features like user-centric, service-oriented and commercial, and present workflow scheduling algorithms are mostly based on the QoS constraints. Many of those algorithms reference the scheduling strategy in grid computing environment. Due to the own characteristics of workflow in a cloud computing environment, the original scheduling policy may have a lack of efficiency in some scenarios.This paper has done a research on workflow scheduling strategy in cloud environment. At first, the thesis introduces a variety of existing workflow scheduling algorithms, and make a analysis of those algorithms'flaws when being applied for scheduling instance-intensive cloud workflows. And then, the thesis introduces the workflow scheduling architecture in a cloud environment, analyzes the business process of a cloud workflow modeled by DAG. Motivated by the problem that the efficiency of existing workflow scheduling algorithms for scheduling instance-intensive cloud workflows is not very high, a QoS constraint scheduling algorithm was proposed, which can minimize the total costs while meeting the deadline. This new algorithm is different from the previous scheduling algorithms that mostly focus on a single complex scientific instance. Then, the CloudSim simulation tool is extended by simulating a DAG database, which randomly generates instances, overriding key classes and some other methods. After that the platform can support the workflow scheduling simulation under the cloud. Finally, the workflow scheduling algorithms is implemented. The experimental results demonstrate that the proposed scheduling approach is more valid for scheduling instance-intensive workflows.