Continuous Performance Benchmarking for Vitess
Talkat KubeCon + CloudNativeCon + Open Source Summit China 2021 · on Friday, 2021-12-10· at 11:20· Kubecon + CloudNativeCon 演讲厅, 虚拟会议 (Virtual)·
summary
Vitess teams run a continuous benchmark system called AreWeFastYet that spins a fresh Equinix M2 xlarge server per run, provisions it with Terraform and Ansible, exercises clusters with sysbench (TPCC/OLTP), scrapes Prometheus metrics and ships them to InfluxDB and MySQL (stored in a Vitess cluster), and posts Slack regression alerts. The site shows concrete comparisons — e.g. ~5.3% OLTP and ~3.4% TPCC gains between v10→v11 — and a dedicated gen4 vs v3 page for planner impact.
AI-written summary · about
chapters (11)
- 0:00Start
- 1:05Concepts
- 1:20Vitess Architecture basics
- 3:49Vitess Architecture Summary
- 4:03New and Upcoming Features
- 4:58Benchmarking
- 7:09What all do we benchmark?
- 7:50Slack Integration
- 8:27Gen-4 Planner - What is it?
- 17:10Implementation of arewefastyet
- 18:22Execution pipeline
related talks
09mGraduated Project Lightning Talk: Vitess, are we Fast Yet?
45mDeep Dive: Vitess
06mProject Lightning Talk: Vitess in Production: Powering Applications at Massive Sca...
40mVitess: Introduction And New Features
40mVitess: Intro and New Features
41mScaling Databases with Vitess
44mIntro: Vitess
36mHorizontal Scaling with Vitess