What is the meaning of RSS(Resident Set Size) of the output of ps
and prstat?
RCLM
RCLM
It's the amount of RAM actually used by the process - as opposedQuote:> What is the meaning of RSS(Resident Set Size) of the output of ps
> and prstat?
--
Rich Teer
President,
Rite Online Inc.
Voice: +1 (250) 979-1638
URL: http://www.rite-online.net
> It's the amount of RAM actually used by the process - as opposed
> to that which has been paged (swapped) out. For example, a 10 MB
> process might have an RSS of 8 MB; this means that 8 MB of the
> process' virtual memory is in RAM, and the other 2 MB is paged
> out to disk.
Like a "limit resident" instead of "limit datasize"?
>Like a "limit resident" instead of "limit datasize"?
On 4.0 this was still present but non-functional.
The reason is that Sun introduced a new orthogonal memory management concept
with SnOS-4.0 and this did not allow to find an 'owner' for a page from
the page info (that is used by the pager daemon).
In 1988 I hacked the scheduler for SunOS-4.0 and found a way to make it work
agin. It was a simple hack. If the scheduler checked a process to decide whether
it should be the next to run, I compared the RSS size with the limit.
If the process was using more than the limit I called swapout() and thus the
process was less 'nice' than other processes. This worked like a charm
for background processes!
Unfortunately, Sun decided to remove the apropriate resource limit with
Solaris 2 aka SunOS-5.x
--
URL: http://www.fokus.gmd.de/usr/schilling ftp://ftp.fokus.gmd.de/pub/unix
AFAIK, no, there isn't.Quote:> Any way to limit the max RSS (quota) on a per process basis,
> without limiting the process's size (swap usage)? If sometimes
IF they're higher priority, they'll be running more often anyay.Quote:> allowed to accumulate a lot of RAM, transfer back to the higher
> priority processes (user's desktop session) when they need it is
> too slow.
If priority paging doesn't do the trick, I can only think of oneQuote:> Like a "limit resident" instead of "limit datasize"?
--
Rich Teer
President,
Rite Online Inc.
Voice: +1 (250) 979-1638
URL: http://www.rite-online.net
To understand why this would be difficult and not very useful, remember
that RSS includes pages of shared objects loaded into the process,
so a processes RSS can increase because a completely unrelated process
loaded in one of the pages in the shared object they both have loaded.
On the other hand, it sounds like what you are looking for is greater
control over how resources are allocated between processes, which is
available via the Solstice Resource Manager product:
http://www.sun.com/software/resourcemgr/
--
________________________________________________________________________
Working for, but definitely not speaking for, Sun Microsystems, Inc.
> To understand why this would be difficult and not very useful, remember
> that RSS includes pages of shared objects loaded into the process,
> so a processes RSS can increase because a completely unrelated process
> loaded in one of the pages in the shared object they both have loaded.
Stabbing at an improvement: Apply the quota instead to a new count,
AccessedSetSize, and defer counting each loaded shared object page towards
any process's "ASS" until they access it first time. ;)
I'll check it out. Thanks for the encouraging responses. Nice toQuote:> On the other hand, it sounds like what you are looking for is greater
> control over how resources are allocated between processes, which is
> available via the Solstice Resource Manager product:
> http://www.sun.com/software/resourcemgr/
1. xterm: huge process size and resident set size
Solaris 9 FCS with Recommended patches through approx. October 2002;
CDE.
What causes an xterm to hang for on the order of a minute when I
resize it? The system is banging noisily on the disk the whole time. I
have approx. 720 MB of free swap out of a GB, 320 MB RAM, and / is
less than 50% full. This is an Ultra 10/ 300 MHz, but I have seen
similar behavior on many other systems and at other installations, so
I'm convinced that it's either an xterm or a Solaris issue.
The scroll-back log is not even that large (although I always start my
xterms with "-sl 10000" (ten thousand lines max scrollback).
I see that the xterm process size is 145M and RSS is 134M, so that
seems to be the immediate cause of the performance issue. The question
then becomes, what could cause those numbers to be so large?, because
most of my other xterms have sizes around 20MB and RSSs of around 3MB.
Thanks for any suggestions.
--Bruce
2. Problems with Netscape in RH 6.0
4. Help! Can't install XWindows
5. resident set size vs. real memory
6. ftp installing of Redhat linux
7. ssh-agent's resident set size increase 8K after each authentication
9. Resident set size on Solaris
10. Swap space and resident set size
11. Solaris: resident size of the process
12. Process resident & total size
13. how to limit resident size of process