hackertube.io

"How NOT to Measure Latency" by Gil Tene

Strange Loop Conference · talk · Watch on YouTube

Software performance & optimizationTesting, debugging & developer tools

What it is

Gil Tene of Azul Systems explains common pitfalls in measuring latency and response time, including why averages mislead and the problem of coordinated omission. He introduces open source tools like HdrHistogram for better measurement and reporting.

Why Hacker News recommends it

The canonical recommendation whenever percentiles, tail latency, load testing or averages come up, described by some as life changing.

  • 9 people recommended it
  • 9 different months
  • 4 strong recommendations
  • first 2015, last 2026

What HN commenters said

  1. “Gil Tene has an excellent talk[1] on this.”

    Excellent talk on worst-case latency and high percentiles

    vikiomega9, 2016-08 · Strong recommendation · in “Throughput vs. Latency and Lock-Free vs. Wait-Free”

  2. “this great talk from Gil Tene on how NOT to measure latencies (basically, don't use averages!).”

    Great talk on why averages mislead for latency

    Rafuino, 2019-04 · Strong recommendation · in “Averages Can Be Misleading: Try a Percentile (2014)”

  3. “This a life changing video by Gil Tene on this topic.”

    High percentiles matter because one customer request fans out to many

    wmfiv, 2020-04 · Strong recommendation · in “Ask HN: Must Read from ACM Library?”

  4. “this talk by Gil Tene is fantastic if you're interested in load testing stats”

    Fantastic talk on load testing statistics

    leetrout, 2023-05 · Strong recommendation · in “Grasshopper – An Open Source Python Library for Load Testing”

  5. “Gil Tene has an entire talk about this point called "How NOT to measure latency”

    Explains why users often hit rare bad-case latencies across dependencies

    telotortium, 2017-02 · Recommendation · in “Unlearning descriptive statistics”

  6. “Worth watching is How NOT to Measure Latency”

    Talk by the author of HdrHistogram on latency measurement

    latch, 2018-10 · Recommendation · in “We can do better than percentile latencies”

  7. “is another really good explanation of this topic”

    Another really good explanation of percentile pitfalls

    allanrbo, 2020-03 · Recommendation · in “Lies, Damned Lies, and Averages: Perc50, Perc95 Explained for Programmers”

  8. “See Gil Tene's talk on how not to measure latency.”

    Argues tail percentiles are what customers actually experience

    agallego, 2020-05 · Recommendation · in “Kafka Removing Zookeeper Dependency”

Excerpts from Hacker News comments, linked to the originals. Summaries and labels were written by a language model from those comments.

Topics: coordinated omission, percentiles, tail latency, hdrhistogram, performance testing

Also recommended by the same people