performance test

이론

• Little’s law by John D. C. Little in 1961
  • Number of concurrent user = TPS * avg. response time
  • Concurrency = TPS * Avg. Response time or Concurrency = TPS * Max. Response time
  • 수용가능한 동시 사용자 = 가상사용자 * Think Time / 평균응답속도T
  • Tomcat MaxThreads = Concurrency * (100% + buffer %)
• Performance anxiety by Joshua Bloch
  • abstraction gap 때문에 성능은 직접 측정해야 알 수 있다
  • In java, run VM 30 times to get meaningful data.
  • http://dl.acm.org/citation.cfm?id=1297033
  • 통계적으로 접근해라.
  • 벤크마크는 잘 못하기가 쉽다.

성능로깅

• http://metrics.codahale.com
• https://code.google.com/p/javasimon
• https://code.google.com/p/parfait
• https://code.google.com/p/stajistics
• https://github.com/Netflix/Hystrix

도구

• https://github.com/tsenart/vegeta

• http://code.google.com/p/caliper/

• http://labs.carrotsearch.com/junit-benchmarks-tutorial.html

  • Junit 기반. H2DB로 결과 저장. 그래프 그려줌.

• http://code.google.com/p/junitbench/

• http://code.google.com/p/caliper/

• Apache benchmark : http://httpd.apache.org/docs/2.2/programs/ab.html

• http://gatling-tool.org/

Jmeter

• Apache JMeter를 이용한 WebSphere 퍼포먼스 테스트

• Jmeter를 사용한 Stress Test(1)

• Jmeter를 사용한 Stress Test(2)

• Jmeter를 사용한 Stress Test(3)

• 사람을 위한 자동화: 전혀 귀찮지 않은 로드 테스팅 (Automation for the people: Hands-off load testing )

• Hands-off Load Testing with JMeter and Ant

• [리뷰 프로파일러 비교]

• 실용적인 성능테스팅

Grinder

• http://grinder.sourceforge.net/

• http://track.sourceforge.net/analyzer.html

• http://code.google.com/p/grinderstone/ (grinder를 Eclipse 내 에서)

• http://sourceforge.net/projects/webflange/ (continuous testing),

• http://wiki.hudson-ci.org/display/HUDSON/Grinder+Plugin

• http://grinder.sourceforge.net/

Locust

• https://cheese10yun.github.io/locust-part-1/