Research On Performance Analysis And Optimization Of Electronic Comerce System Based On Partial Servers Vacation

It is necessary to make performance analysis and optimization on ElectronicCommerce System (ECS) in order to improve its quality of service and decrease itsoperational cost. Hence, the research on theory and technology of ECS performanceanalysis is realistic and urgent, and the research achievement can be not only directlyapplied to ECS performance analysis but also be used and referenced in the performanceanalyses on Web service, computer network system, distributed system. Some of theservers may go to do assistant task at idle time. This random absence is often called asvacation. Based on vacation queue theory and guided by the route map given by author,an ECS performance model has been built, and the performance metrics are presented,and their mathematics solutions have been given, and a flexible analysis and optimizationplatform has been designed.Firstly, the essence of EC is presented for ECS performance analysis, and a logiclayer model of transmitting and processing information in ECS was described, key itemseffecting ECS performance were analyzed. In order to solve workload unbalance and tomake performance analysis in ECS, vacation queue policies were introduced into ECSoperation. The closed property of conditional probability forΓ distribution classes wasproved, which plays an key role in mathematics analysis of performance model.Secondly, the logic models of queue in ECS were depicted in detail, the relationformulas between the transition probability of user behavior and the transition probabilityof request among servers was defined. Several queue models were unified on the modelsof partial servers vacation, and applied scenarios were given. Using quasi-birth-and-deathprocess, the mathematics analysis models were accurately built. Applying matrixgeometric solution, the analyses and solutions were made on the partial servers vacationmodels M/M/c such as single, multiple, N-policy, asynchronous vacation, and mathematicexpressions of stationary waiting queue length and waiting time were obtained, and theircondition stochastic decomposition properties were established. In this way, theperformance metrics of ECS can be calculated and the performance comparison betweenvacation queue model and classic queue model can be easily made.Furthermore, a route map on ECS performance analysis has been given and theperformance analysis can be carried out in 5 phases. The ECS performance metrics onbusiness, usability, client and server were summarized. The course of ECS performanceanalysis is concretely illuminated by an ECS example, in which the performance metricsin different scenarios such as response time, queue length, waiting time etc. were solvedout, and their comparisons were made. The new concepts, vacation service capacity,vacation-delay rate and vacation probability, were brought forward, and their mathematicexpressions were defined, and they also applied in decision of optimized vacation policyand the number of vacation servers.Finally, to apply the research results easily and directly, and to solve the problems ofperformance analysis software that brought by user demand change and environmentchange, the concepts of software flexibility such as flexible degree, flexible force, flexibledistance etc. are given. Flexible software technologies such as flexible report, flexibleinformation issue have been put forward, and applying them a flexible platform of ECSperformance analysis and optimization has been designed.