Abstract: Aiming at the application context for data stream constrain,this paper introduces a tool that composes the web services with centralized model,then takes the algorithm to partition the centralized code into decentralized codes for decentralizing execution.The authors have applied multilevel algorithms for graph partitioning to partition the centralized code in order to satisfy the constrain of minimizing the data stream between the domains and maximizing the throughput of the distributed system.During the process of partitioning,the task of minimizing the edge-cut can be considered the objective and balancing the multi-resource load can be considered as the constraint.The instance analysis indicates that the algorithm can accomplish the task automatically and quickly,and furthermore balance the multi-constraint load of the distributed system.

Key words: graph partition, web service composition, data stream constrain, load balance

摘要： 在WebJet平台工具支撑下,针对数据流约束的应用环境,对Web服务采取集中式组合,分布式运行的技术路线。在满足划分之间数据流量最小化以及分布式系统吞吐量最大化的目标约束下,运用图划分的多级算法将Web服务组合划分成分布式代码片段。在划分过程中,以极小化边割权值和为目标,多资源负载均衡为约束条件。实例分析说明该算法能自动、快速地将集中式BPEL程序划分为分布式的BPEL程序,并对BPEL程序迁移到的分布式节点之间的负载进行基于多资源约束的均衡调整,使分布式运行系统达到负载均衡。

关键词: 图划分, Web服务组合, 数据流约束, 负载均衡