FW: FW: LRU's and CLEANERS (IDS 9.21FC1)

FW: FW: LRU's and CLEANERS (IDS 9.21FC1)

Post by Dirk Moolma » Tue, 06 Feb 2001 17:56:49



-----Original Message-----


Sent: Saturday, February 03, 2001 4:39 PM

Subject: Re: FW: LRU's and CLEANERS (IDS 9.21FC1)



>Thanks for the input Christian.  I just still have to be convinced about
>this one.  I think your suggestion is also in line  with what the manual
>says, however I've heard of a lot of  people having 128 LRUS  on only a
>couple of cpus.   In a previous post by Art Kagel (Wed 11/29/2000 6:19 PM),
>he mentioned that when you experience bufwaits (onstat -p), you should up
>the LRUS.

>If I follow the manual, I will have to decrease my LRUS, if I follow Art's
>advice, I will have to increase my LRUS. I think I'll try to increase them
>for now and then take it from there.

>  Increase! Monitor the effect with onstat -g spi

>  with only 8 lRUS you should find that the second column becomes
>  very high for lru-?? entries!

Thank you David, just one more question, what would be considered a high
value under
"onstat -g spi" for the lru-entries ?

Here is a sample of ours -
Num Waits   Num Loops   Avg Loop/Wait    Name
6869        258573      37.64            fast mutex, lru-0
311         22370       71.93            fast mutex, lru-1
6599        244812      37.10            fast mutex, lru-2
334         21092       63.15            fast mutex, lru-3
6686        239281      35.79            fast mutex, lru-4
278         24724       88.94            fast mutex, lru-5
6415        205450      32.03            fast mutex, lru-6
347         28934       83.38            fast mutex, lru-7

>Thanks again
>Dirk
>Reach Technologies

>"People demand freedom of speech as a compensation for the freedom of
>thought which they seldom use."
>- Kierkegaard

>-----Original Message-----


>Sent: Friday, February 02, 2001 11:38 AM

>Subject: Re: LRU's and CLEANERS (IDS 9.21FC1)



>> The manual says the following about LRU's and CLEANERS

>> To monitor the percentage of dirty pages in LRU queues, use the
>onstat -R
>> command. When the number of dirty pages consistently exceeds the
>> LRU_MAX_DIRTY limit, you have too few LRU queues or too few page
>> cleaners. First, use the LRUS parameter to increase the number of
>LRU
>> queues.
>> If the percentage of dirty pages still exceeds LRU_MAX_DIRTY, use
>the
>> CLEANERS parameter to increase the number of page cleaners.

>> My parameters have the following values -
>> LRU_MAX_DIRTY  2
>> LRU_MIN_DIRTY  1
>> LRUS  30
>> CLEANERS  30

>> onstat -R gives me
>> 8587 dirty, 498979 queued, 500000 total, 524288 hash buckets, 2048
>buffer
>> size
>> start clean at 2% (of pair total) dirty, or 333 buffs dirty, stop at
>1%
>> 0 priority downgrades, 0 priority upgrades

>> According to what the manual says above, I need to increase my LRU's
>and
>> CLEANERS.  Is it still true if max & min dirty is set to 2 & 1
>respectively
>> ?  To make a long story short, how many LRU's and CLEANERS do I need
>?

>At SAP we recommend to configure LRUS and CLEANERS after NUMCPUVPS:

>        LRUS = max( 4, NUMCPUVPS + 2 )
>        CLEANERS = LRUS + 2

>After our experience we find performance advantages over the
>"traditional" settings, especially with low LRU_MAX/MIN_DIRTY setting
>and KAIO.

>HTH
>Christian
>--
>#include <std_disclaimer.h>

--
David Williams

 
 
 

1. FW: LRU's and CLEANERS (IDS 9.21FC1)

Thanks for the input Christian.  I just still have to be convinced about
this one.  I think your suggestion is also in line  with what the manual
says, however I've heard of a lot of  people having 128 LRUS  on only a
couple of cpus.   In a previous post by Art Kagel (Wed 11/29/2000 6:19 PM),
he mentioned that when you experience bufwaits (onstat -p), you should up
the LRUS.

If I follow the manual, I will have to decrease my LRUS, if I follow Art's
advice, I will have to increase my LRUS. I think I'll try to increase them
for now and then take it from there.

Thanks again
Dirk
Reach Technologies

"People demand freedom of speech as a compensation for the freedom of
thought which they seldom use."
- Kierkegaard

2. Can't update data control

3. LRU's and CLEANERS (IDS 9.21FC1)

4. Data modifications/Querying a linked Server (2) Questions

5. FW: FW: FW: FW: Explain (more is better?)

6. SQL Agent - Online Pending

7. FW: FW: FW: Same configuration ?

8. Filemaker, ODBC and MS Access

9. FW: FW: FW: Explain (more is better?)

10. Fw: [HACKERS] Re: Fw: Isn't pg_statistic a security hole - Solution Proposal

11. FW: FW: FW: FW: FW: Explain (more is better?)

12. FW: IDS 9.30UC2 failed to run with RH-7.3/glibc-2.2.5-39

13. Fw: [GENERAL] Fw: Please Help