how to know process working set size in linux /proc

Question

Process Working Set Info in LINUX

I am finding Working Set Size of process in proc folder this link say that I can find working set size in /proc but I don't know how to know. I knew RSS is working set size but RSS is different from working set size can I know working set size using /proc/[pid]/statm?

Answer 1

RSS (resident virtual size) is how much memory this process currently has in main memory (RAM). VSZ (virtual size) is how much virtual memory the process has in total.

From your question I believe you're after virtual size, i.e. total memory usage.

Regarding statm - from Linux manpages:

proc/[pid]/statm
Provides information about memory usage, measured in pages.  The columns are:

size       (1) total program size
           (same as VmSize in /proc/[pid]/status)
resident   (2) resident set size
           (same as VmRSS in /proc/[pid]/status)
share      (3) shared pages (i.e., backed by a file)
text       (4) text (code)
lib        (5) library (unused in Linux 2.6)
data       (6) data + stack
dt         (7) dirty pages (unused in Linux 2.6)

So you need the first integer, which will return total page count used. If you however need in more readable output, status will provide information in kilobytes. For example:

rr-@burza:~$ cat /proc/29262/status | grep -i rss
VmRSS:      1736 kB
rr-@burza:~$ cat /proc/29262/status | grep -i vmsize
VmSize:     5980 kB

This means process 29262 uses 5980 kB, out of which 1736 resides in RAM.

Answer 2

I don't believe that /proc/[pid]/statm gives the WSS, or /proc/[pid]/status for that matter.

WSS is the number of pages a process needs in memory to keep "working".

RSS is the number of pages of a process that actually reside in main memory.

So RSS >= WSS. Meaning that RSS may include some pages that the process doesn't really need right now. Maybe it used those stale pages in the past.

From my understanding of linux internals, the kernel doesn't really keep track of the WSS on a per-process basis. WSS is too involved to track continuously and doesn't have an exact formula. RSS is simpler to calculate, so the kernel just reports that.

Note that if the sum of WSS of all processes is greater than or equal to the main memory size (i.e. the system is thrashing or close to thrashing) then RSS equals WSS because only the pages absolutely needed by a process are kept in the main memory. Got it?