Veebirakenduste tehingute tõenäosuslik testimine
Date
2011
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Tartu Ülikool
Abstract
IT süsteemid muutuvad oma elutsükli vältel järjest keerulisemaks. Veebirakendusi kasutatakse eriti laialt erinevatel eesmärkidel, sest võrgupõhine juurdepääs informatsioonile on väga mugav. Kuid võrgupõhise juurdepääsu juures tekivad mõned probleemid, mida tuleks silmas pidada. Kasutajad eeldavad prognoositavat jõudlust (nt nõuetekohane reaktsiooniaeg), seega teenusepakkujad peavad teadma, kuidas nende süsteem töötab erinevate koormuste all. Selles teesis loome tõhususe analüütilise mudeli ja töötame välja programmi, mis selle lahendab. Antud programm lubab analüüsida veebirakenduste jõudlust ja vastata järgmistele küsimustele:
1)missugune on keskmine süsteemi reaktsiooniaeg?
2)missugune on süsteemi kasutamine üldiselt?
Parameetrid programmi jaoks nagu keskmine teenindusaeg, uute taotluste keskmine saabumisaeg, keskmine mõtlemisaeg, on saadud testsüsteemi reaalse koormuse logidest. Jõudluse mudel on välja töötatud Queuing Networksi abil, mis lubab analüüsida süsteemi matemaatiliste valemite abil.
Web systems are used widely for reaching different purposes, as remote access to information is very convenient. However, the remote access brings many aspects which should be handled. Users expect predictable performance levels (e.g., acceptable response time), therefore, service providers should know how their system performs under different loading conditions. In this thesis I design an analytical performance model and develop a tool which can solve that model. The tool allows analyzing the performance of web applications and answer the following questions: 1)What is the average response time of the system? 2)What is the utilization of the system as a whole? The input parameters, such as the average service time of transactions, average arrival rate of requests, and the average think time, are estimated from a real workload (of a system under test). The performance model is developed by means of Queuing Networks, a framework which enables the analysis of a system in terms of mathematical formula.
Web systems are used widely for reaching different purposes, as remote access to information is very convenient. However, the remote access brings many aspects which should be handled. Users expect predictable performance levels (e.g., acceptable response time), therefore, service providers should know how their system performs under different loading conditions. In this thesis I design an analytical performance model and develop a tool which can solve that model. The tool allows analyzing the performance of web applications and answer the following questions: 1)What is the average response time of the system? 2)What is the utilization of the system as a whole? The input parameters, such as the average service time of transactions, average arrival rate of requests, and the average think time, are estimated from a real workload (of a system under test). The performance model is developed by means of Queuing Networks, a framework which enables the analysis of a system in terms of mathematical formula.