Chasing that 32 ms tail latency in ClickHouse
From the source: ClickHouse Blog
A customer saw an extra ~32 ms on simple SELECT queries after idle periods. We traced it to unbatched Native protocol metadata packets overflowing a reset TCP congestion window, reproduced it with network namespaces, and fixed it by batching them.
Read the full story on ClickHouse BlogAsk the square
The story is what was announced. If it raises a question for your team, ask the people who run this technology day to day. Your post links back to this story, so the next reader finds it too.
Recent threads in Databases
What is your approach to zero-downtime schema migrations at scale?
Expand-and-contract, online schema change tools, feature flags? Share what has kept large-table migrations boring.
Postgres everywhere vs purpose-built databases: where do you draw the line?
Vector search, time series, queues, analytics: many teams now start with Postgres extensions. Where has that worked, and where did you need a dedicated engine?
Something wrong or out of date? Suggest a correction. We reply within 5 business days.
More from ClickHouse
Recent updates from ClickHouse, so you can tell whether this is a one-off or part of a pattern.
More in Databases
What other companies in Databases are doing. The category page shows who’s active, side by side.
