JVM profileerimise jõudluse optimeerimine Honest Profileri baasil
Abstract
Honest Profiler on tööriist, mis võimaldab mõõta Java virtuaalmasina peal jooksvate rakenduste jõudlust. Tööriista poolt kogutud informatsiooni põhjal on võimalik optimeerida vaadeldava rakenduse jõudlust. Käesoleva töö eesmärk on luua lahendusi, mis suurendaksid Honest Profileri tööriista poolt kogutud informatsiooni hulka. Suurem andmete hulk muudab jõudluse mõõtmise tulemused täpsemaks. Töö kirjeldab profiilide kogumise ning Honest Profileri arhitektuuri põhitõdesid. Ühtlasi mõõdetakse Honest Profileri informatsiooni kogumise loogika jõudlust. Töö põhitulem on kolm erinevat lähenemist, mis suurendavad kogutud informatsiooni hulka. Kirjeldatud lahenduste jõudlus ning kogutud informatsiooni hulk verifitseeritakse jõudlustesti abil.
Honest Profiler is a profiling tool which extracts performance information from applications running on the Java Virtual Machine. This information helps to locate the performance bottlenecks in the application observed. This thesis aims to provide solutions to increase the amount of useful information extracted by Honest Profiler. Achieving this would increase the accuracy of the performance information collected by Honest Profiler. Thesis will cover the basics of sampling profiling, the architecture of Honest Profiler and measures the performance of Honest Profiler’s data collection logic. As the main result of this thesis, three different solutions for increasing the profiler information output are presented. Their performance and the extracted information amount is evaluated by a benchmark test.
Honest Profiler is a profiling tool which extracts performance information from applications running on the Java Virtual Machine. This information helps to locate the performance bottlenecks in the application observed. This thesis aims to provide solutions to increase the amount of useful information extracted by Honest Profiler. Achieving this would increase the accuracy of the performance information collected by Honest Profiler. Thesis will cover the basics of sampling profiling, the architecture of Honest Profiler and measures the performance of Honest Profiler’s data collection logic. As the main result of this thesis, three different solutions for increasing the profiler information output are presented. Their performance and the extracted information amount is evaluated by a benchmark test.