A Performance Modeling And Optimization Approach Based On Deployment Scheme

Software develops towards the direction of a kind of dynamic,open and distributed style gradually.SaaS software which provides cloud services in the Internet generated from then on.However,SaaS software performance degradation problems constantly occur due to business logic structures of the problems or excessive external access in the process of SaaS software run-time.How to improve software performance faces a deal of significant challenges.SaaS software performance optimization which is deployment scheme oriented is an important approach to help make software performance better by adjusting the mapping relationships between software services and virtural machines in the cloud.This dissertation,focusing on the characteristics of SaaS software and the specific requirements of software performance optimization,proposes a method of performance modeling and optimization based on deployment scheme oriented.The main contributions are as follows:(1)This dissertation proposes SSDL,a SaaS Software Deployment Language supporting performance optimization and constructs software performance model.SSDL considers software configuration informations,performance requirements and resource availability requirements according to the characteristics of SaaS software deployment and information SaaS software performance optimization requires.This dissertation constructs the conceptual model of SSDL language describing SaaS software deployment scheme from four aspects which are deployment environment,deployment object,deployment plan and deployment constraint.The work of this paper is based on the existed models and methods and can achieve the following design requirements:(a)abstract the process of software deployment and define sufficient language elements;(b)provide necessary information for SaaS software performance optimization and support the implementation of the following redeloy work;(c)apply to dynamic deployment environment and extend conveniently.(2)This dissertation proposes a kind of software performance optimization approach based on deployment scheme.SaaS software deployment scheme performance optimization is a combinatorial NP-Hard problem.We often face the problems such as huge search spaces,enormous QoS history data and so on in the process of finding the most optimized mapping relationships between services and virtual machines.Firstly,I construct a performance optimization input function according to deployment object and deployment environment information of SSDL,and then I build a performance optimization constraint function on the basis of deployment constraint information of SSDL.I adopt linear regression to construct performance optimization objective function and compute to acquire deployment plan after optimization through linear programming method in the end.(3)This dissertation takes an experiment using an actual SaaS software system in order to verify this approach.This paper enforces the performance modelling and optimization with an urgent task processing management system.Firstly,I describe the deployment scheme of the system using SSDL language,and then I construct the necessary input function,objective function and constraint function.Finally,an improved deployment scheme is acquired after simulation using Lingo.It has been demonstrated that the approach this dissertation puts forward can model and optimize software performance based on deployment scheme.Therefore,this approach can help system deployment staff maintain the normal running of softwares.