Company
Date Published
Author
James Katz
Word count
1118
Language
English
Hacker News points
None

Summary

Heap faced performance challenges with its multi-petabyte Postgres database cluster due to rapid data growth during COVID, which led to storage issues and performance degradation. The company identified that the utilization of SSDs near capacity, especially with the ZFS filesystem's copy-on-write (CoW) operations, was causing the slowdown. To address this, Heap explored upgrading to ZFS 2.x with Zstandard compression, which promised a better compression ratio and significant disk space savings compared to the existing lz4 compression. Rigorous testing revealed that the switch to Zstandard maintained query performance while reducing storage usage by 20% and halving average write durations. Consequently, Heap rolled out the new configuration across its database nodes, resulting in substantial storage savings and improved write performance, demonstrating the impact of minor configuration adjustments at scale and the value of leveraging open-source software solutions.