LRU, Cleaners and num disks
Bridget Reitsma asked for a formula relating LRUs and page cleaners to the number of disks and buffers. Yuri Dovgart suggested cleaners at least equal to chunks and ideally at least the LRU count, and about 1 LRU per 500-750 buffers. Another poster gave a worked example with 45000 buffers, 45 LRUs and 15 cleaners. Watawinona advised at least 4 LRUs, independent of disks, and onstat -R to judge cleaners. No formula.
Auto-generated by Claude from the posts below — may be imperfect; read the full thread.
Hi, Does some one out in informix land have a magic formula of working out what the raito of LRU's and cleaners should be? Is is dependant on the number of disks? I read that there should be one cleaner per disk but they should not exceed 50 - this seems a little over the top. Also I understand that the LRU's are dependant on buffer size - has any one got a formula to work this out too? I know no formula's are 100% correct but I need some where to start before I tune my engine again. cheers bridget
Bridget Reitsma <b.reitsma@auckland.ac.nz> wrote in message news:38FCCB86.D88D17A1@auckland.ac.nz... > Hi, > > Does some one out in informix land have a magic formula of working out > what the raito of LRU's and cleaners should be? Is is dependant on the > number of disks? Sometimes. First of all number of cleaners must be at least equal to the number of chunks (because this is optimal for checkpointing). From the other side it will be good, if number of cleaners will be equal to the number LRUs, or will be higher. Remember, that having too much cleaners will take too much CPU time, when having cleaners too low will cause your system to slower checkpointing and LRU writes. I read that there should be one cleaner per disk but > they should not exceed 50 - this seems a little over the top. > > Also I understand that the LRU's are dependant on buffer size - has any > one got a formula to work this out too? I think 1 LRU per 500-750 buffers will be good. > > I know no formula's are 100% correct but I need some where to start > before I tune my engine again. Administrators Guide ... :). -- ------------------------------------------------- With best regards, Yuri Dovgart SAP R/3, Informix technical consultant, Informix Certified Professional, Senior System Consultant System Architecture and High Availability Systems, 'Telecominvest' company Email y_dovgart@tci.ukrtel.net ICQ 39284285
In article <38FCCB86.D88D17A1@auckland.ac.nz>,
Bridget Reitsma <b.reitsma@auckland.ac.nz> wrote:
> Hi,
>
> Does some one out in informix land have a magic formula of working out
> what the raito of LRU's and cleaners should be? Is is dependant on
the
> number of disks? I read that there should be one cleaner per disk but
> they should not exceed 50 - this seems a little over the top.
>
> Also I understand that the LRU's are dependant on buffer size - has
any
> one got a formula to work this out too?
>
> I know no formula's are 100% correct but I need some where to start
> before I tune my engine again.
>
> cheers bridget
>
>
There are various parameters related:
BUFFERS 45000 # Maximum number of shared buffers
CLEANERS 15 # Number of buffer cleaner processes
LRUS 45 # Number of LRU queues
LRU_MAX_DIRTY 3 # LRU percent dirty begin cleaning limit
LRU_MIN_DIRTY 1 # LRU percent dirty end cleaning limit
BUFFERS/LRUS = Size LRU # 1000 in the example (2Mb, Buffer size=2Kb)
As LRU_MAX_DIRTY=3% and LRU_MIN_DIRTY=1% a Page cleaner will move 2Mb
for one LRU
As CLEANERS=15 each page cleaner will clean 3 LRUS (Will move 6 Mb max.)
I hope this example be useful for you, I don't know any magic formula,
sorry
Sent via Deja.com http://www.deja.com/
Before you buy.
>Subject: LRU, Cleaners and num disks
>From: Bridget Reitsma b.reitsma@auckland.ac.nz
>Date: 18.04.00 22:54 W. Europe Daylight Time
>Message-id: <38FCCB86.D88D17A1@auckland.ac.nz>
>
>Hi,
>
>Does some one out in informix land have a magic formula of working out
>what the raito of LRU's and cleaners should be? Is is dependant on the
>number of disks? I read that there should be one cleaner per disk but
>they should not exceed 50 - this seems a little over the top.
>
>Also I understand that the LRU's are dependant on buffer size - has any
>one got a formula to work this out too?
>
>I know no formula's are 100% correct but I need some where to start
>before I tune my engine again.
>
>cheers bridget
>
>
>
>
>
>
>
keep the number of LRUs at at least 4. You can increment up to the number of
CPUs and beyond if a lot of contension occurs. The number of LRUs is
independent of the number of disks, and onstat -R will tell you if you need
more CLEANERS.
Nona