Queueing Theory Based Service Replica Strategy for Business Process Efficiency Optimization in Community Cloud

Abstract

Community Cloud (CC) emerges as a rapid developing paradigm in many fields such as research and education. Business process (BP) that facilitates the organization cooperation and resource sharing plays an important role in CC. The execution efficiency of BP has a pivotal impact on the quality of community cloud services. CC applications often have high demands on BP execution efficiency, which is affected by many factors such as response time of a single service in BP and cost of interactive activities among distributed services. This brings up an important research issue of optimizing BP execution efficiency for CC. In this paper, we propose a service replica strategy that utilizes queuing theory to optimize the execution efficiency of BP for CC. Our approach creates service replicas dynamically to shorten the response time of request to a single service according to the quantity and frequency of user requests in different times, and allocates these replicas to proper host servers to reduce the interaction time among distributed services. Our objective is to improve the BP execution efficiency by the above two aspects of local time optimization. First, we analyze the effects of service replica strategy on optimizing BP execution efficiency as a whole and define a cost model of execution efficiency to reduce the search space of maximizing execution efficiency of BP. Then we implement the replica strategy in two steps. We initially utilize the social network property of the services to pre-allocate key service replicas on server nodes in CC. After that, we utilize a queueing model to schedule the service replica and service requests dynamically. Experiments illustrate that our approach can help to improve BP execution efficiency in CC.