多网络环境下基于爬山聚类算法的SOA性能优化

Optimizing SOA performance in multi-network environment based on hill-climbing clustering algorithm

为了提升多网络环境下面向服务的架构(SOA)的性能,提出基于频度权重连接强度(CSWPF)和部署空间的网络性能模型,并给出对应的爬山聚类算法完成部署优化.CSWPF以单位时间内网络服务间的数据交互量作为系统性能度量;部署空间则根据业务规则确定每个网络服务在不同子网内部署的灵活性.在部署空间的约束下,爬山聚类算法以CSWPF作为度量尺度,通过不断尝试各种网络服务的部署方案,降低网络间流量,应用现有技术提升性能.仿真实验表明,算法在获得或者逼近最优解方面有较高的效率.项目实践表明,该方法可以明显降低系统负荷,提升性能.

A networking performance model and the hill-climbing clustering algorithm for deployment optimization were introduced to improve the service oriented architecture （SOA） performance based on the connectivity strength with parameter and frequency （CSWPF） and the deployment space.CSWPF provides the performance measurement by calculating the communication cost per unit time;and the deployment spaces define Web services＇ deployment possibility in different sub-networks according to the business logic.In the deployment spaces,the hill-climbing clustering algorithm tries different possible deployment solutions to achieve the object of inter-network flow reduction based on CSWPF.Simulation shows that the nearly best optimization solution can be obtained quickly.Project experiment shows that this algorithm can improve the system performance by reducing the network traffic.