TechPlanetScaleabout 17 hours ago
Concurrency vs. Throughput: why more parallelism can make databases slower
9 min read
A PlanetScale production MySQL database melted down for sixteen minutes after a batch job held row locks for fifteen minutes, causing ten thousand concurrent snapshot reads to fail. Vitess's transaction pool was raised to ten thousand, worsening backpressure. Reconfiguring vttablet to mimic a thread pool eliminated errors and kept QPS flat during a later burst.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
