Code-Memo

Disk performance tuning

Measure first

Identify what’s slow:

• latency vs throughput
• random vs sequential IO
• read vs write

Quick checks:

lsblk -o NAME,TYPE,SIZE,ROTA,MOUNTPOINTS
df -h

Observe IO

If installed:

iostat -xz 1 5

Per-process:

pidstat -d 1

Common levers

• Filesystem choices: ext4 vs xfs vs btrfs (workload dependent)
• Mount options: e.g. noatime
• IO scheduler: depends on device type and kernel defaults
• Read-ahead: can help sequential workloads

Caching and memory

• More RAM can reduce IO (page cache)
• Don’t starve the system; avoid aggressive cache drops in production

SSD vs HDD

• HDD: sensitive to random IO and seeks
• SSD/NVMe: much better random IO; watch write amplification and endurance

Related

• System monitoring