Why does ext4 perform periodic cpu-based work, and how can these cpu interruptions be minimized
Issue
-
Systems with ext4 filesystems periodically interrupt cpu-based workloads to perform filesystem-related maintenance functions. This can sometimes result in undesirable cpu latencies for high-performance, latency-sensitive user applications. What background work is being performed in the filesystem code, and how can these latencies be reduced and/or eliminated?
-
In particular, relatively large cpu spikes have been observed at approximately 30-second intervals.
- Smaller spikes are also observed at 25ms intervals.
Environment
- RHEL 6
- ext4 filesystems
- latency-sensitive applications executing on isolated cpus
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.