
Why Database Compaction Can Stall Under Write Load
On this episode of Database Tech with Fexingo, Lucas and Luna dig into a database problem that quietly wrecks performance: compaction. When a storage engine based on LSM trees merges sorted runs in the background, it can suddenly steal CPU and disk bandwidth, causing latency spikes and write stalls. The hosts explain why compaction is necessary, how it can go wrong, and what engineers can do to keep it from ruining their p99s. They walk through the classic trade-off between write amplification and read performance, and discuss real-world strategies like leveled compaction, size-tiered compaction, throttling, and manual compaction windows. If you've ever seen a database slow down for no apparent reason, this episode is for you.
#DatabaseCompaction #LSMTree #StorageEngine #WriteAmplification #PerformanceTuning #DatabaseInternals #Engineering #DataEngineering #Technology #FexingoBusiness #BusinessPodcast #FexingoShow #SQL #NoSQL #DatabaseTech #Latency #BackendEngineering #Infrastructure
Keep every episode free: buymeacoffee.com/fexingo
Información
- Programa
- FrecuenciaCada día
- Publicado30 de agosto de 2026 a las 10:44 p.m. UTC
- Duración10 min
- Temporada4
- Episodio171
- ClasificaciónApto