topic Re: RSS vs VSS in Operating System - HP-UX

RSS x VSS

Wagner Villela — Fri, 16 Mar 2012 17:54:47 GMT

Dear friends,

Any way to get use of RSS memory by a process with not via glance? I think "ps" gets information about VSS use by an process and it comes with some shared area included, right? I try:

# UNIX95= ps -e -o pid,sz,pcpu,ruser,args

PID        SZ      %CPU RUSER    COMMAND
29622    2601    0    avb2020 oracleceeec
19029    24     0        avb2012 AVB_pcserv
19785   2513   0    avb2018 oracleceeec
28402 23      0        avb2010 AVB_pcserv
28621 2491     0      avb2004 oracleceeec

Colunm "SZ" gets VSS information and not RSS information, right?

Thanks for all

Re: RSS vs VSS

Dennis Handly — Fri, 16 Mar 2012 19:19:15 GMT

>Column "SZ" gets VSS information and not RSS information?

It's hard to be sure with ps(1). It mentions sz and vsz:

sz The size in physical pages of the core image of the process, including text, data, and stack space.

vsz The size of the process in (virtual) memory in kilobytes (1024 byte units).

But vsz does sound like it gets VSS. Whether that means sz is RSS, I'm not sure.

Re: RSS vs VSS

Wagner Villela — Fri, 16 Mar 2012 22:17:34 GMT

@Dennis Handly wrote:
>Column "SZ" gets VSS information and not RSS information?

It's hard to be sure with ps(1). It mentions sz and vsz:
sz The size in physical pages of the core image of the process, including text, data, and stack space.
vsz The size of the process in (virtual) memory in kilobytes (1024 byte units).
But vsz does sound like it gets VSS. Whether that means sz is RSS, I'm not sure.

Thanks, Dennis.

I'm still searching for more details.

Glance information is ok, but i need to sum all process of a given user and output format of "ps" is perfect for this need. Unfortunately seens that information is in VSS sizes.

Re: RSS vs VSS

Dennis Handly — Sat, 17 Mar 2012 05:04:48 GMT

>I need to sum all process of a given user. Unfortunately seems that information is in VSS sizes.

From the source, the SZ and VSZ info comes from:

struct __pst_status {
SZ:
        _T_LONG_T pst_dsize;    /* # real pages used for data
        _T_LONG_T pst_tsize;    /* # real pages used for text
        _T_LONG_T pst_ssize;    /* # real pages used for stack

VSZ:
        _T_LONG_T pst_vtsize;   /* # virtual pages used for text
        _T_LONG_T pst_vdsize;   /* # virtual pages used for data
        _T_LONG_T pst_vssize;   /* # virtual pages used for stack
        _T_LONG_T pst_vshmsize; /* # virtual pages used for shared memory
        _T_LONG_T pst_vmmsize; /* # virtual pages used for mem-mapped files
        _T_LONG_T pst_vusize;   /* # virtual pages used for U-Area & K-Stack
        _T_LONG_T pst_viosize; /* # virtual pages used for I/O dev mapping
        _T_LONG_T pst_vrsesize; /* # virtual pages used for RSE stack

top: _T_LONG_T pst_rssize; /* resident set size for process (private pages)

Re: RSS vs VSS

Wagner Villela — Sat, 17 Mar 2012 10:20:25 GMT

@Dennis Handly wrote:
>I need to sum all process of a given user. Unfortunately seems that information is in VSS sizes.

From the source, the SZ and VSZ info comes from:
struct __pst_status {
SZ:
        _T_LONG_T pst_dsize;    /* # real pages used for data
        _T_LONG_T pst_tsize;    /* # real pages used for text
        _T_LONG_T pst_ssize;    /* # real pages used for stack

Dennis

You mean that the information from "ps" and showed in "SZ" column is in fact a RSS memory use?

Is a bit confused to me, because if i run a "ps" and do a sum of all values of "SZ" column, the final result is much more than physical memory of the machine.

By the way, this machine works with a very, very high load and perform swap mostly part of time. This total of "SZ" that exceds physical memory may be because machine does swap?

Thank you very much

Re: RSS vs VSS

Dennis Handly — Sat, 17 Mar 2012 10:45:05 GMT

>You mean that the information from "ps" and showed in "SZ" column is in fact a RSS memory use?

Well, it isn't, why have both:

_T_LONG_T pst_dsize; # real pages used for data

_T_LONG_T pst_vdsize; # virtual pages used for data

>if I run a "ps" and do a sum of all values of "SZ", the final result is much more than physical memory of the machine.

The text size is shared.

>this machine works with a very, very high load and perform swap mostly part of time.

Then why bother measuring it if it is that bad? ;-)

This total of "SZ" that exceeds physical memory may be because machine does swap?

That and shared text and using VSS may explain it.

Re: RSS vs VSS

Wagner Villela — Sun, 18 Mar 2012 20:51:51 GMT

@Dennis Handly wrote:
>You mean that the information from "ps" and showed in "SZ" column is in fact a RSS memory use?
Well, it isn't, why have both:
_T_LONG_T pst_dsize; # real pages used for data
_T_LONG_T pst_vdsize; # virtual pages used for data

>if I run a "ps" and do a sum of all values of "SZ", the final result is much more than physical memory of the machine.

The text size is shared.
>this machine works with a very, very high load and perform swap mostly part of time.

Then why bother measuring it if it is that bad? ;-)
This total of "SZ" that exceeds physical memory may be because machine does swap?

That and shared text and using VSS may explain it.

Thanks, Dennis.

If you remember some magic way to sum all RSS values of running process, please let me know.

Thanks again.

Re: RSS vs VSS

Dennis Handly — Mon, 19 Mar 2012 01:14:53 GMT

>magic way to sum all RSS values of running process

You can redirect the top output with "-f file" and scan that.

Re: RSS vs VSS

Wagner Villela — Mon, 19 Mar 2012 12:09:52 GMT

@Dennis Handly wrote:
>magic way to sum all RSS values of running process
You can redirect the top output with "-f file" and scan that.

Hummm.....

top -d 1 -n 1024 -f lix1

Output works fine. But, again, in "top" manual we can read:

**** about RES column ****

RES

Resident size of the process in kilobytes. The resident size information is, AT BEST, AN APPROXIMATE VALUE.

Any way to know if it can represent a big difference in values?

Thanks again. Very useful tip.

Re: RSS vs VSS

Dennis Handly — Wed, 21 Mar 2012 04:56:42 GMT

>Any way to know if it can represent a big difference in values?

I assume by approximate is that it doesn't account for shared regions?