Presentation: "How not to measure latency"
Time: Wednesday 13:30 - 14:20
Location: Regency
Understanding application responsiveness and latency is critical to delivering good application behavior. But good characterization of bad data is useless. When measurements of response time present false or misleading latency information, even the best analysis can lead to wrong operational decisions and poor application experience.
In this talk, Gil Tene (CTO, Azul Systems) will discuss some common pitfalls encountered in measuring and characterizing latency. Gil will demonstrate and discuss some false assumptions and measurement techniques that lead to incorrect results, and cover simple ways to sanity check and correct these situations. He will discuss the fallacy of using standard deviation measurements, the strongly multi-modal nature of latency, common discontinuities found in most computing platforms, and how back pressure and coordinated data omission issues can skew measurement results dramatically. Gill will introduce and demonstrate how simple and recently open sourced tooling can be used to improve and gain higher confidence in both latency measurement and reporting.