isn't buf cache dynamic on 11.11?

[Sergey Akifiev_1]

question is in the subject.
isn't buf cache supposed to shrink if system is running out of memory? on 11.11.
and on 11.23?

[Peter Godron]

Sergey,
from the often quoted:
ftp://hprc.external.hp.com/memory.htm#review_buffer_cache_size

"To check the current size of the buffer cache, either "sysdef | grep bufpagesâ (and multiply by 4096 to approximate the current size of buffer cache) or use glanceâ s memory screen to see what size â BufCacheâ is.

Note: Although buffer cache can be dynamic in size, decrease only occurs under memory pressure and then only decreases very slowly. So, the buffer cache often grows fairly quickly to dbc_max_pct and only decreases (and slowly) when memory presssure is high."

[Sergey Akifiev_1]

and what is meant to be "pressure" in this case? system totally runs out virtual memory (phys mem + swap)?

[Peter Godron]

Sergey,
servers can and do run out of virtual memory. Release, as stated, is a slow process !

Looking at your previous questions via your profile, few of your questions have been answered to your satisfaction as they remain open. Can you please update them with the level of answer you would like to see.

[Sergey Akifiev_1]

i don't uderstand, what do you mean:
"Can you please update them with the level of answer you would like to see"

should i close those threads?

[Peter Godron]

Sergey,
what are your dbc_min_pct and dbc_max_pct settings , as they will also impact on the release. See http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=70315

My apologies regarding my comment on your ealier threads:
1093702:
Has your update completed ? Did Dave's answer fix the problem?
1091588:
Your initial question on how to check async was answered? Can you please close.
1085251:
No update since 20/12/06.

General:
http://66.34.90.71/ITRCForumsEtiquette/after.html

[Bill Hassell]

Yes, the buffer cache will shrink as more and more programs occupy more and more RAM. Now a program can request a lot of local or shared memory and get the infamous "program too big" or ENOMEM (errno 12), or "out of memory" messages, but this is seldom caused by a buffer cache that has not shrunk down.

Virtual memory on HP-UX means that once RAM + swap has been completely used, only then will the system be out of memory. However, a 32bit program will get the error when it tries to request more memory than the MAXDSIZ kernel parameter will allow, or more memory than the 32bit addressing method will allow. So the kernel will (and has since 10.01) properly reduce the cache size as additional RAM is used by processes. The real reason for "out of memory" messages requires more investigation.

[Sergey Akifiev_1]

we have intentionally leaved dbc_max_pct and dpc_min_pct to their defauilt values to see if buffer cache is really dynamic.

and now we have 10-40MB RAM free and about 5-10 de/reactivations per sec. also about 130MB of swap used.
nevertheless buffer cache size only decreased by 100M and now it is about 1.4B.

as for other my topics, well, they're still unanswered. so i have nothing to post there more.

[Bill Hassell]

Attached is a program to grab a *lot* of memory. When you run it, it asks how much RAM you'd like to use. I would recommend grabbing about 250 megs each time. The internal comments show you how to compile it in several different ways. Because it actually writes a pattern into the requested memory, it will bypass reservation and actually use the swap area once it runs out of RAM. In 64 bit mode, the program can request all of your RAM and all of your swap space, though not in a single call. And you'll see the buffer cache shrink rapidly. At 11.23, the change is far faster than 11.11.

[Bill Hassell]

Oopos, wrong program. mallocmax only requests the maximum amount of RAM. The attached mallocmegs will actually write into the RAM area.

[Sergey Akifiev_1]

thnx, Bill. i'll try that on some test system.