Study On Scalable Cloud Workflow Engine And Its Implementation

With the development of service-oriented compute pattern,more and more cloud service is springing up based on cloud computing.And workflow technology can coordinate and combine various services,realizing complex business logic in a reusable way.It is receiving more and more attention in cloud computing.Now workflow has becoming an IT service in companies and even a public service towards various companies.At the same time,to lower the cost and keep stabilized performance,workflow can also be deployed on public cloud,using the advantages of cloud computing like on-demand billing,flexibility and so on.While such cloud workflow service uses the advantages of cloud computing to achieve better cost performance,it is also faced by a lot of new problems.Firstly,by providing service in cloud environment,the system may receive large amount of requests,the workflow engine may probably need distributed computing to handle these requests.Secondly,with the increase of the requests,the fluctuation of requests is also increasing.Fixed amount of compute resource may lead to a waste of compute resource.The workflow engine will need an appropriate control algorithm to dynamically adjust the compute resource being used,based on the system's demand.Thirdly,the territory selectively of compute resource in cloud computing also brings many new subjects for workflow.By using compute resource on different location,we can optimize the call time of web services on various location.Based on these problems,we design and implement a scalable cloud workflow engine,trying to apply workflow technology on cloud environment and provide cloud workflow service.In this paper,the main research work and innovative results reflected in the following aspects:1.Design and implement a scalable clod workflow service,with management and monitoring platform of the service system.The cloud workflow service can execute cloud workflows,and the management platform realizes the load balancing,monitoring,dynamic adjustment functions for the whole system.2.Design and implement a self-adaptive control algorithm.The algorithm can adjust the cloud resources used by system based on the history and current system load,making the system's performance able to satisfy the dynamic load and try to avoid waste.3.Study on cloud workflow distributed execution problem,and proposed an approximation algorithm.It calculates each activity' execution region in in advance,and then allocate them into different region to reduce the delay time in communication and decrease the workflow execution time.