vmstat

Vmstat, short for virtual memory statistics, is a performance monitoring utility for Unix-like systems. It reports a compact summary of the system’s memory usage, process activity, CPU utilization, paging, block I/O, interrupts, and system calls. The data come from kernel counters and can help diagnose memory pressure, swap behavior, CPU saturation, and I/O bottlenecks. Vmstat can run once or in a continuous loop to observe trends over time.

Usage and output

Typical invocation is vmstat [delay [count]]. If a delay is supplied, vmstat prints a header followed by repeated samples every delay seconds, for count times if provided. Without a delay, it prints a single snapshot. The output is divided into columns grouped as procs, memory, swap, IO, system, and CPU. Procs shows r (runnable) and b (blocked). Memory lists swpd (amount of virtual memory used), free, buff (buffers), and cache. Swap includes si (swap in per second) and so (swap out per second). IO shows bi (blocks in from I/O devices) and bo (blocks out). System reports in (interrupts per second) and cs (context switches per second). CPU aggregates user, system, idle time, with optional fields for IO wait and virtualization stolen time on some platforms (us, sy, id, wa, st).

Options and variability

Common options include -a (active/inactive memory), -d (disk statistics), -f (forks since boot), -s (summary since boot), and -p (per-device statistics). The exact options and the precise column labels can vary across Unix variants and kernel versions, so behavior may differ between Linux, BSD, and other systems.

Use cases

Vmstat is frequently used for quick health checks, baseline measurements, and in scripts to monitor trends or trigger alerts related to memory pressure, paging activity, or I/O contention. For deeper analysis, it is often complemented by tools like top, iostat, sar, or perf.