Event-based Distributed Workflow Technology In Cloud Environment

The cloud service platform BirisCloud is the research background of thisdissertation.Service is provided on demand by BirisCloud which is a SaaS-basedfourth-party service platform.Users, providers,resources and the service centers areintegrated to offer service.Fistly,process models are used to describe the businessprocess by service center,and then the workflow running nodes are rented to supportthe execution of its business process according to the process definition.In order toestablish the relationships between process models,event-driven mechanism isused,and because of the many-to-many relationship of service centers and workflowrunning nodes,the distributed workflow technology is used to accomplish the taskallocation and the accurate scheduling of workflow engines.In this dissertation,on the level of basic knowledge and theories of workflowtechnology,the JPDL process definition language and the method of establishing thegrammatically correct,rational and reliable process models by using the workflownet and the reachability graph as well as the process and principle of creation andexecution of process instance are researched.On the basis of elaborating the cloud service platform BirisCloud,requirementanalysis from the angle of operation and user is made in this dissertation. Taking theevent-driven mechanism and distributed workflow as the core of research directionis clear.Since all process models are treated as atomic process models,the relationshipbetween process models are established depending on events.Based on this,theevent-driven mechanism is emphatically studied in this dissertation.Taking thedefinition,formal description and classification of event and event instance asstarting point, event publishing and subscription mechanism is researched andproposed by this dissertation,with which the tightly coupled relationship betweenprocess models can be decoupled. After the relationship between process models areestablished by events,the ECA-based event-driven rules, the operation principle ofevent-driven distributed workflow,the production and consumption mechanism ofevent instance and the core algorithm are researched to support the analysis of the event instance at the operation phrase,so as to achieve the purpose of scheduling ofworkflow engine correctly.In order to make the workflow tasks running correctly in the case of themany-to-many relationship between service centers and workflow nodes,theimplementation of distributed workflow on the process model level and theworkflow engine level is researched.Three kinds of distributed workflowarchitectures which are suitable for the cloud service platform BirisCloud areproposed.And the comparison analysis on stability, robustness and response timeamong the three architectures are made.Furthermore, the scheduling problem ofdistributed workflow is researched and the dynamic load balance schedulingalgorithm is proposed,with which the scheduling of workflow engine in distributedenvironment is optimized.At last,as the verification and support for the theoretical research, event-baseddistributed workflow management system in cloud environment is designed anddeveloped.The complete case analysis and verification of the system is made,whichis from the angle of the establishment,operation and monitoring of the serviceprocess of multi-restaurant service center.