Blazin’ Fast PromQL
Talkat KubeCon + CloudNativeCon North America 2019 · on Tuesday, 2019-11-19· at 10:55· Room 11AB - San Diego Convention Center, Upper Level·
summary
Prometheus’ single-threaded PromQL engine limits large queries, so Cortex builds a query front-end that does step alignment, per-day partial queries and result caching (memcached for immutable chunks, careful index caching). Splitting a 7-day query into seven parallel subqueries cuts 2.5s → 1.4s without cache and to ~0.2s with cache; a heavy dev query drops ~10s → ~3s with caching+parallelism (aiming for 100ms at millions of series). Front-end also works in front of Prometheus/Thanos.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:22Cortex
- 3:58Prometheus
- 14:21Demo
- 18:21Whats next
- 19:45Aggregation
- 20:26MapReduce
- 23:08Parallel Processing
- 23:56Future Plans
- 25:25Questions
- 31:53Metrics
related talks
29mInside a PromQL Query: Understanding the Mechanics
29mReverse-Engineering PromQL Usage: A Proxy’s Tale
01h 19mIntro + Deep Dive: Prometheus
45mScaling Prometheus with Cortex, one time series at a time
29mBeyond TSDB: Unlocking Prometheus with Parquet for Modern Sc...
32mPrometheus Intro and Deep Dive
39mHow To Take Prometheus Planet Scale: Massively Large Scale Metrics Installations
23mAutomatically Making Dashboards Load 100X Faster