Use vmstat to Monitor System Performance

What is vmstat?

Vmstat is a built-in Linux system monitoring tool. Its primary job is measuring a system’s usage of virtual memory. No matter how powerful it is, a Linux server has a finite amount of random access memory (RAM). A Linux system can run out of RAM for several reasons, such as demands on the operating system and its running applications. When this happens, the Linux kernel swaps or pages out programs to the computer’s storage devices, called swap space. Typically, this is a reserved area of hard drive or solid-state drive storage. It’s used as virtual memory when RAM is unavoidable. As RAM is freed up, the swapped-out data or code is swapped back into the main RAM-based memory.

System performance drops drastically when swapping happens. That’s because the server’s swap I/O speed is much slower than RAM even if the hardware – such as fast SSD – is used for virtual memory. In addition, when Linux uses virtual memory it spends more of its CPU cycles on managing virtual memory swapping.

That means that Linux system administrators have to pay attention to a server’s memory usage. Nobody wants a system to slow down. When it does, and admins need to troubleshoot, virtual memory is a likely culprit. Since virtual memory has a big impact on system performance, vmstat is essential for monitoring it. In addition to monitoring virtual memory paging, vmstat also measures processes, I/O, CPU, and disk scheduling.

How to Use vmstat

You can run vmstat both as an interactive program and in shell programs. When you run vmstat without any parameters, it shows system values based on the averages for each element since the server was last rebooted. These results are not a snapshot of current values.

Run vmstat using the following command:

    vmstat

You see a similar output:

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 1171312  69752 2231152    0    0     0    13   10   23  0  0 100  0  0

Notice how vmstat returns information about the system’s processes, memory, swap, input and output, system interrupts and context switches, and CPU. To learn more about each column and value, view the manual pages for vmstat by issuing the man vmstat command and searching for each specific area. For example, once you are viewing the vmstat manual pages, to view more information about the swap column, issue the /swap command.

The vmstat command-line options provide more information about your system. The vmstat syntax is as follows:

    vmstat [options] [delay] [count]]

Options: vmstat command settings
Delay: the time interval between updates. If no delay is specified, the report runs as an average since the last reboot.
Count: the number of updates printed after the given delay interval. If no count is set, the default is an infinite number of updates every x seconds (where x = delay).

For example, you can issue the following command to run vmstat every five second, five times on an idle system:

    vmstat 5 5

Your output should resemble the following:

procs -----------memory---------- ---swap-- -----io---- -system-- ------cpu-----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
 1  0      0 1166396  70768 2233228    0    0     0    13   10   24  0  0 100  0  0
 0  0      0 1165568  70776 2233352    0    0     0     8  121  224  0  0 99  0  0
 0  0      0 1166608  70784 2233352    0    0     0    53  108  209  0  0 100  0  0
 0  0      0 1166608  70784 2233352    0    0     0     0   75  176  0  0 100  0  0
 0  0      0 1166576  70788 2233352    0    0     0     4   76  177  0  0 100  0  0

vmstat Commands

vmstat is often run with an interval of 1 second for a small number of seconds depending on kind of problem the administrator is trying to diagnose. The following example illustrates an interval of one (1) second twenty (20) times:

vmstat 1 20