How to Identify Linux Bottlenecks with vmstat: CPU, Memory, I/O Explained

When a Linux system shows a high load average, you must determine whether the bottleneck is CPU, memory, or I/O. The vmstat command provides a concise snapshot of these resources.

1. View system load with vmstat

procs -----------memory---------- -----swap-- -------io---- --system-- ------cpu-------
r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
0  0      0 496056 889316 4065748    0    0     9    41   55   51  0  0 99  1  0

Procs (r b)

r : Number of processes running or waiting for a CPU time slice. If this stays above 1 for a long period, the CPU is likely insufficient.

b : Number of processes blocked for resources such as I/O or memory swapping.

Memory (swpd free buff cache)

swpd : Amount of memory swapped to disk (in kilobytes). A non‑zero or large value (e.g., >100 MB) may indicate memory pressure, but if si and so remain zero, performance can still be acceptable.

free : Amount of free memory (in kilobytes) currently available.

buff : Memory used as buffer cache, typically for block device I/O.

cache : Memory used as page cache for the filesystem. A large cache suggests many cached files; if I/O bi is low, the filesystem is operating efficiently.

Swap (si so)

si : Amount of memory swapped in from disk per second.

so : Amount of memory swapped out to disk per second.

I/O (bi bo)

bi : Blocks read from block devices per second (KB/s).

bo : Blocks written to block devices per second (KB/s). If bi + bo exceeds 1000 KB/s and wa is high, consider balancing disk load, possibly using iostat for deeper analysis.

System (in cs)

in : Number of interrupts per second.

cs : Number of context switches per second. A high cs relative to I/O and network rates warrants further investigation.

CPU (us sy id wa st)

us : Percentage of CPU time spent in user mode. Values consistently above 50 % may indicate the need to optimize user processes.

sy : Percentage of CPU time spent in kernel mode. If us + sy exceeds 80 %, the CPU is likely a bottleneck.

wa : Percentage of CPU time spent waiting for I/O. Values above 30 % suggest serious I/O wait, possibly due to random disk access or limited disk bandwidth.

id : Percentage of CPU idle time.