how to get cache hit percent for aix

On Dec 6, 11:53 pm, volkan <avyildi...@gmail.com> wrote:[color=blue]
> Hi Gurus,
>
> I'm stuck...Just as I feel like I'm understanding the vmo parameters,
> I'm getting confused again...The problem(question) raised from the
> situation that, after getting two banks merged,and after losing the
> openview admins, the new management decided not to use the openview
> anymore and wanted me to merge the systems monitored by openview, to
> the system monitor application they've been using, which is a well-
> written C code, but for solaris systems only...I have a huge number of
> aix servers, so more or less, I reached a point that I mostly port
> everything to the new monitoring environment...
>
> The points which I cannot decide/understand what to do with, are:
>
> 1) Openview performance reporter is perfectly plotting the performance
> data, and the chart I mostly use while
> monitoring the performance is the "memory summary"...In this, a line
> is called "cache hit percent"...This data is
> very important as you all know in order to understand what's going on
> with the performance...The problem is,
> I couldn't produce or calculate the similar "cache hit percent"
> outputs with openview by using any of the commands/libraries I know...
> sar -b gives a report but I guess it is meaningless, because it always
> shows zero for every field, which is quite an absurd output...
> While looking at the man mage of sar, in the -b, it says:
>
> -b Reports buffer activity for transfers, accesses, and cache (kernel
> block
> buffer cache) hit ratios per second. Access to most files in Version 3
> bypasses
> kernel block buffering and therefore does not generate these
> statistics
>
> which would be the reason of meaningless outputs...Clearly OVO looks
> at something else, but I couldn't find a way...Not in the output of
> svmon, or in the libperfstat.h...
>
> Does anyone know how to get the "real" cache hit percentage (for every
> second, let's say) for aix? In solaris, sar -b reports very well, and
> besides we have the dtrace for everything...
>
> 2) While I was dealing with the question above, and trying to find a
> solution, I came to a point where I was fighting with the vmo
> parameters, maxfree, minfree, maxperm, minperm, strict_maxperm,
> etc...So, nmon, or let's say vmstat -v, shows
> [root@opcbaswprd]</usr/include>vmstat -v
> 3145728 memory pages
> 3024067 lruable pages
> 2194231 free pages
> 3 memory pools
> 343618 pinned pages
> 80.0 maxpin percentage
> 3.0 minperm percentage
> 8.0 maxperm percentage
> 7.8 numperm percentage
> 238879 file pages
> 0.0 compressed percentage
> 0 compressed pages
> 7.8 numclient percentage
> 8.0 maxclient percentage
> 238879 client pages
> 0 remote pageouts scheduled
> 2 pending disk I/Os blocked with no pbuf
> 0 paging space I/Os blocked with no psbuf
> 2484 filesystem I/Os blocked with no fsbuf
> 0 client filesystem I/Os blocked with no fsbuf
> 2536 external pager filesystem I/Os blocked with no
> fsbuf
> 0 Virtualized Partition Memory Page Faults
> 0.00 Time resolving virtualized partition memory page
> faults
>
> As far as I understand, the "file pages" and the "numperm percentage"
> lines show information about the size of the memory being used by the
> cache...What I don't understand is, man pages say that xxperm
> parameters are for JFS, while for JFS2, we have to deal with
> xxxclient parameters...So I'm a little confused here, does the
> "numclient percentage" show the percent of jfs2 pages being cached,
> and "numperm percentage" the jfs ones? If so, how can I have both of
> the outputs, where my system has all-jfs2 filesystems?
> Any helps to remove my confusion would welcome :)[/color]

A part of answer to your query is that numperm includes ALL file pages
where as numclient shows the percentage of JFS2, NFS and GPFS pages.
In your case as shown in the output of vmstat numperm - numclient is 0
which means that there are no JFS pages on your system! HTH