一种基于性能模型的中间件自配置框架

高动态的计算环境使得QoS(quality of service)保障对于基于组件的分布式系统越来越重要,软件系统需要具备自我调整的能力以适应外部环境的变化.给出一种自适应的中间件配置框架,能够动态感知负载变化,并自动调整系统参数配置以保持用户所要求的服务质量.该框架的核心是一个基于分层排队网络的性能预测模型,用于指导搜索最优的资源配置,使性能需求得到最大的满足.在OnceAS应用服务器上进行原型实现,并以StockOnline应用做实验,比较了在使用和不使用该框架时的性能需求的满足情况.结果显示,在负载增加时,通过自配置框架的调控,应用性能需求的保障程度得到了较大的提升.

A Performance Model-Based Self-Configuration Framework for Middleware

High dynamic computing environment makes QoS(quality of service)guarantee more important for component-based distributed system.Software system should possess self-tuning capacity for reacting to external environment variation.This paper proposes an adaptive self-configuration framework,which can automatically tune configuration parameters to preserve QoS as workload changes.The key of this framework is a layered queuing network based performance model,and it guides the search for the best combination of configuration parameters to satisfy the QoS requirement.This self-configuration framework is prototyped on OnceAS application server,and is validated using StockOnline by comparing the performance requirement satisfaction with and without this framework.The results show that through the framework's regulation,system performs well on QoS goal.