Performance testing with machine learning

Performance testing with machine learning

Доклад затрагивает такую важную часть, как нагрузочное тестирование. В докладе будет рассмотрены процессы по внедрению в пайплайн, уменьшению ручной работы по анализу и подготовке данных, а так же применение машинного обучения для анализа результатов по части квантилей.

405cf5a7e7cad100e431fa3a2403e909?s=128

Zoya Chizhkova

October 24, 2019
Tweet

Transcript

  1. 1.

    Align Technology, Inc. 1 © 2019 Align Technology, Inc. 1

    Confidential – For Internal Use Only New part of pipeline Performance testing with machine learning
  2. 2.

    Align Technology, Inc. 2 Agenda • Basic problems of performance

    testing • Why did we decide to move forward? • Architectural solution • What have we improved already? • What are we going to do after?
  3. 4.

    Align Technology, Inc. 4 Daily routine of performance engineers ENV

    x (S + DU + RG + RA) = DT Test dependency environments test data Test results results analyzing results gathering Build dependency scaling BD x DT = ?
  4. 6.

    Align Technology, Inc. 6 Scaling and data Monitoring System Performance

    test Get test variables Load for requests ENV config Cache Service Test data
  5. 8.

    Align Technology, Inc. 8 Build CI Run script Test run

    MS Docker container Gathering Analyzing PASS/FAIL Tests results
  6. 10.

    Align Technology, Inc. 10 Jmeter Logs Service Custom Service Interface

    of java class from custom JAR Client side as 90% percentiles Server Monitoring Service Generate Results Server side as 90% percentiles Errors from logs Send metrics for visualization
  7. 13.

    Align Technology, Inc. 13 We mark a build as passed

    90% percentiles from client side and server side are ok There are no errors from logs service There are no degradation from DB
  8. 15.

    Align Technology, Inc. 15 DB check from the client side

    with python plots Postgre connections HikariPool connections