Statistics reported by svmon are expressed in terms of pages. A page is a 4K block of virtual memory while a frame is a 4K block of real memory.
These parameters show the number of both the working and persistent pages in RAM.
svmon -G shows global memory reports (in 4K page sizes)
size inuse free pin virtual
memory 4096000 2423149 1672851 803943 1734262
pg space 2097152 103815
work pers clnt other
pin 665509 0 99 138335
in use 1711933 0 711216
PageSize PoolSize inuse pgsp pin virtual
s 4 KB - 2284685 103815 731591 1595798
m 64 KB - 8654 0 4522 8654
memory
size - size of real memory in 4k pages (in kbytes=4096000 * 4)
inuse - size of used memory in 4k pages (total memory minus free)
free - amount of memory on the free list in 4k pages (free RAM)
pin - amount of memory pinned in RAM in 4k pages (pinned memory is always resident in RAM and cannot be paged out)
virtual - memory needed to execute the programs on the system (for non-filesystem cache) (same as avm in vmstat)
pg space
size - size of paging space
inuse - number of paging space pages used
pin (subset of real memory containing pinned pages (665509+0+99+138335=803943))
work - number of working pages pinned in RAM (processes ...)
pers - number of persistent pages pinned in RAM (JFS)
clnt - number of client pages pinned in RAM (JFS2/NFS/GPFS)
other - could not find the meaning :(
in use (subset of real memory in use (1711933+0+711216=2423149))
work - number working pages in RAM (processes ...)
pers - number persistent pages in RAM (JFS)
clnt - number client pages in RAM (JFS2/NFS/GPFS)
PageSize (details of 4KB 64KB 16MB pages) (2286685+16*8654=2423149)
Inuse - size of used memory of the given PageSize pages
Pgsp - size of used paging space of the given PageSize pages
Pin - size of pinned memory of the given PageSize pages
virtual - size in memory for non-fs cache of the given PagesSize pages
In the below case 885*16MB is reserved but nothing is using it (inuse=0)
PageSize PoolSize inuse pgsp pin virtual
s 4 KB - 4289065 9348 804589 3562062
m 64 KB - 10536 0 8102 10536
L 16 MB 885 0 0 885 0
This could be checked with vmo as well:
vmo -a|egrep "lgpg_regions|lgpg_size"
lgpg_regions = 885
lgpg_size = 16777216 <-16MB
IBM recommendation was to turn this off:
vmo -r -o lgpg_size=0 -o lgpg_regions=0 <--we did a reboot after that
The large (16MB) and supreme pages (16 GB) need to be configured manually, while the small (4k) and medium (64K) sizes are preconfigured by default. When you install an AIX system, the small and medium pages will be active. How these page sizes are used is determined by AIX. When a process needs memory AIX will assign automatically the best memory page size regarding small and medium pages. The biggest performance advantage that large memory pages have over the small- or medium-sized pages is that they’re nonpageable. This means they’re "pinned" pages, which are exempt from page outs to disks.
------------------
------------------
------------------
svmon -P 278620
-------------------------------------------------------------------------------
Pid Command Inuse Pin Pgsp Virtual 64-bit Mthrd LPage
278620 shm_lgpg64 13781 11319 1083 13769 Y N Y
PageSize Inuse Pin Pgsp Virtual
4 KB 5589 3127 1083 5577
16 MB 2 2 0 2
This process uses 13781 pages in real memory. 11319 frames are pinned.
The number of pages reserved or used in paging space is 1083.
------------------
Pid Command Inuse Pin Pgsp Virtual 64-bit Mthrd 16MB
2781436 oracle 14086839 65566 1784541 14122372 Y N N
16MB N: it means this process does not use the 16MB pages, N=no
8 comments:
Hi,
use the below command to find the top 15 process with details.
svmon -Pt15 | perl -e 'while(<>){print if($.==2||$&&&!$s++);$.=0 if(/^-+$/)}'
Hi, thanks for this really nice command :-)
apart from above command you can use svmon -P -O unit=MB|head -15(it will also display same thing in MB)
Except
64-bit,
Mthrd,
16MB
How i can get overall memory and CPU utilization in percentage ?
I'm trying in AIX box..
For memory using = svmon -G -O unit=GB
For cpu using = sar -u 5 5
using script how i can get the value in percentage and schedule as a job
how can you manage high paging size, meaning using command #ipcrm ? I've done a research on the net, and saw people uning it with success for processes with NATTCH equal zero. What's your opinion on that?
Post a Comment