dbc_max_pct

[SSP_1]

Hi Guys ,

Model - N4000-44, Hpux 11.00, Memory - 8.00 GB

We observe a gr8 amount of memory utilisation ~99% for the above server, the average util is 99% and it's for constant basis. Following is the memory status retrived from glance output
Total VM : 1.02gb Sys Mem : 374.0mb User Mem: 3.34gb Phys Mem: 8.00gb
Active VM: 648.6mb Buf Cache: 4.00gb Free Mem: 296.8mb

I understand that if we make the buffers percentage less [right now dbc_max_pct is 50% and dbc_min_pct is 5%] then we can have more room of physical memory available for util. Considering disk util is not much can I change the dbc-max-pct to 25% and dbc_min_pct to 10%.

Pl advise.

[Massimo Bianchi]

Hi,
first question is what are you using this server for.

If you have oracle, then advised parameters are

dbc_max_pct 8
dbc_min_pct 5


In general, since you have 8G of RAM, i would not use more than

dbc_max_pct 15
dbc_min_pct 5

Did you do a check with glance for effective used values ?

HTH,
Massimo

[Michael Tully]

The suggested and recommended dbc_max_pct setting is 300-500MB no more. So in your case setting dbc_max_pct should be 5% as an ideal with the dbc_min_pct being 2%

You'll still have to reboot your system but I would expect a big benefit from doing this.

[SSP_1]

Thanks Masimo,

Would it be feasible to change it to

dbc_max_pct 25%
dbc_min_pct 10%

Checked in glance for used values - What does it mean???

[Massimo Bianchi]

Hi,
dbc-max-pct to 25 is too high, also as Michal pointed out.


glance is a tool for performance monitoring of the system. It's a payment tool, but maybe you bought it.

try with "glance" or "gpm" (graphical version).


If not avaiable, there is a 60 days free trial, you can teste it and see if it satisfies your needs.

With glance you can monitor the used buffer, and decide which value can be best for you.



HTH,
Massimo

[Paula J Frazer-Campbell]

HI

Dbc max and min should be set not by a set percentage but by a calculation.

A good starting point is to set min for circa 80 Meg and max for 300-400 meg

So if you have 8 gig of memory then I would suggest :-

dbc Min - 2%
dbd Max - 10%

And then monitor, if all ok then by calculation Min can go to 1% and Max to 5%

Paula


Paula

[Tom Geudens]

Hi,
For a very good explanation of this (and a lot of others) parameter, check out
http://h21007.www2.hp.com/dspp/files/unprotected/devresource/Docs/TechPapers/UXPerfCookBook.pdf

It's one of my favorites (and written very clearly).

Regards,
Tom Geudens

[SSP_1]

Hi ,

Regarding the same problem , again i found that my swap is pseudo one and the current util is as given below.

root@moses:>swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4196 0 4196 0% 0 - 1 /dev/vg00/lvol23
dev 4196 0 4196 0% 0 - 1 /dev/vg00/lvol25
reserve - 3021 -3021
memory 6352 5313 1039 84%
total 14744 8334 6410 57%

Is it mean that it's using 5 Gb out of 8 Gb for paging only, is this a reason to grow mem util to 99%.

If I get a memory report from glance it's as given below.


Total VM : 1.02gb Sys Mem : 374.0mb User Mem: 3.34gb Phys Mem: 8.00gb
Active VM: 626.6mb Buf Cache: 4.00gb Free Mem: 297.3

[Tom Geudens · Accepted solution]

Hi again,
Go with the glance-report. It shows that you're definitely oversized for buffer-cache, but that user-memory is not even half of the 8Gb. You'll see definite improvement as soon as you alter those buffer-cache parameters.

Use the same document I told you about in my previous answer to read up on pseudo-cache ... but do NOT turn it of (as the document suggests). There are several opinions about this, and in my company we consider wasting a couple of Gigabytes of diskspace on "never-to-be-used" cache to be a sin if you can save it by turning on pseudo-cache.

It's a trade-off really ... and you have to consider it carefully.

Regards,
Tom Geudens

[Tom Geudens]

Hi,
I meant pseudo-swap, not pseudo-cache in my answer.

Sorry about that,
Tom Geudens