Font Size: a A A

Research On Evaluation Mechanism Oriented Services Availability Using Combinatorial Accelerated Stress Testing

Posted on:2016-07-03Degree:MasterType:Thesis
Country:ChinaCandidate:Z B HeFull Text:PDF
GTID:2348330542475450Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
Availability is the key indicator of services provided by computer system and the lifeline of highly available services.In order to provide more credible highly available services,carrying out more accurate and more credible services availability evaluation to realize accurate availability evaluation,or even predicting the influences to services availability by faults or failures have become the urgent priority of computer system or services availability.Web services are the most common and important services in highly available services,and the study of availability evaluation oriented Web services is of great significance to the promotion of users' experience.However,existing researches on evaluation mechanism oriented services availability have several problems,such as testing environment is not real enough,evaluation results are imprecise,unreliable and haven't reached transaction level and so on.Therefore,it's an essential research on combinatorial accelerated stress testing and evaluation mechanism oriented services availability.The paper focuses on improving accuracy and credibility of availability evaluation for Web services in special testing environment.Firstly,we propose a strategy of combinatorial accelerated stress testing(CAST)to simulate actual environmental stress while someone is using Web services.Then,We propose an improved grey correlation analysis availability evaluation mechanism with entropy difference and weight value(EW_GCA),and a confidence validation algorithm.The strategy of combinatorial accelerated stress testing(CAST)which we proposed,aiming at simulating actual environmental stress and fusing stress testing technology and accelerated testing technology,has built an availability testing model combining three stress style including step-up-stress,constant-stress and unprogressive-stress.The paper uses HP LoadRunner,through building environment,designing scenario,demarcating transactions and setting parameters,to implement combinatorial accelerated stress testing for Web services and proceed a SLA analysis according to the simulation experiment results,which has provided more accurate and more credible test sample volumes for availability evaluation.To solve the problems of evaluation results are imprecise,unreliable and haven't reached transaction level in existing researches,the paper also proposes an improved grey correlation analysis availability evaluation mechanism with entropy difference and weight value(EW_GCA)and confidence validation algorithm,this mechanism only needs the maximum and minimum value of each operation or transaction in availability evaluation index set for Web services,through establishment and standardization of availability evaluation interval numbers decision matrix,construction of services optimal vector and suboptimal vector,confirmation of optimal subordinate degree of services etc.,with entropy difference to illustrate the changes of dependability and weight value to quantize dependability requirements of different operations or transactions in network services.Finally,according to the parameter estimation theory in the field of mathematical statistics,we use maximum likelihood estimation,hypothesis testing and confidence coefficient for normal population methods to further validate the availability evaluation results.The experiment result shows that this mechanism can realize hierarchical analysis and overall evaluation for Web services' availability more accurately and more credibly,even in the case of small test sample volumes.
Keywords/Search Tags:Availability evaluation, Web services, Combinatorial accelerated stress testing, Grey correlation analysis, Confidence coefficient
PDF Full Text Request
Related items