TME10 [ITM] Computational memory attribute

[Howell, Joseph]

In AIX you can have memory managed in 4k, 64k, and ridiculously-large pages. In TEPS, there's a "computational memory" display that shows memory use calculated as the number of 4k pages in use (both computational and non-comp). It appears that internally a conversion is being done from the actual number of pages into the equivalent number of 4k pages. Is this true? (I'd look it up in the manual but the ITM Information Center isn't working for me at the moment. I miss paper manuals....except for the revision pages part.)

[cid:image007.jpg-VC6WQkJ+***@public.gmane.org]<http://www.dickssportinggoods.com/home/index.jsp>

JOSEPH HOWELL
SYSTEMS ENGINEER
345 Court Street l Coraopolis, PA 15108
Joseph.Howell-***@public.gmane.org<mailto:Joseph.Howell-***@public.gmane.org> l o: 724.273.4182

[cid:image008.png-VC6WQkJ+***@public.gmane.org]<http://www.facebook.com/dickssportinggoods> [cid:image009.jpg-VC6WQkJ+***@public.gmane.org] <https://twitter.com/DICKS>







The information contained in this message and any attachments (collectively, the "Transmission") from Dick's Sporting Goods, Inc. or one of its subsidiaries contains confidential information and is intended solely for the named recipient(s). If you are not a named recipient, you are prohibited from copying, distributing or using this Transmission. Please contact the sender immediately by returning the e-mail and deleting the original Transmission.

[g***@gmail.com]

Joe: A look in /usr/include/libperfstat.h shows that the
"perfstat_memory_page_t" structure itself, which is where the comp and
non-comp memory are derived, has a a "psize" attribute which is used for
the denominator portion of the math:

.....
* *psize_t psize; /* *page size in bytes*
*/
u_longlong_t real_total; /* number of real memory frames of this
page size */
.....

However, it's not clear how to confirm psize's present value.
libperfstat.h also defines the structure for "perfstat_psize_t" and
includes a metric called "real_avail" which is defined as the "number of
pages (in 4KB pages) of memory", but this explicitly pertains to virtual
(paging space) memory...or perhaps not. Anyway, I agree with you that it
looks like a conversion takes place before or after the structure is parsed
and passed to the Agent by the data provider. If you like to be 100%, L3
can quickly confirm as much via a PMR, making this opinion moot.... :-)

Toben

Post by Howell, Joseph
In AIX you can have memory managed in 4k, 64k, and ridiculously-large
pages. In TEPS, there’s a “computational memory” display that shows memory
use calculated as the number of 4k pages in use (both computational and
non-comp). It appears that internally a conversion is being done from the
actual number of pages into the equivalent number of 4k pages. Is this
true? (I’d look it up in the manual but the ITM Information Center isn’t
working for me at the moment. I miss paper manuals….except for the
revision pages part.)****
** **
<http://www.dickssportinggoods.com/home/index.jsp>****
*JOSEPH HOWELL***
*SYSTEMS ENGINEER*
345 Court Street l Coraopolis, PA 15108****
* *
<http://www.facebook.com/dickssportinggoods> <https://twitter.com/DICKS>**
****
** **
The information contained in this message and any attachments
(collectively, the "Transmission") from Dick's Sporting Goods, Inc. or one
of its subsidiaries contains confidential information and is intended
solely for the named recipient(s). If you are not a named recipient, you
are prohibited from copying, distributing or using this Transmission.
Please contact the sender immediately by returning the e-mail and deleting
the original Transmission.
_______________________________________________
TME10 mailing list