IDS Forum

Re: BUFFERPOOL question

Posted By: Art Kagel
Date: Tuesday, 17 February 2009, at 11:06 p.m.

In Response To: Re: BUFFERPOOL question (jak aka)

No!

The lru_min/max_dirty parameters control how many buffers are flushed
BETWEEN checkpoints, not at checkpiont time. So:

case1 lru_max = 51 , lru_min=50
This means that when any lru is 51% dirty flush it clean to 50%. But at
checkpoint time ALL of the dirty buffers, which in this case will average
50.5% of all buffers (or 505,000 buffers) will have to be flushed making the
checkpoint VERY long and before 11.50 blocking all new transactions and new
portions of existing transactions from continuing.

case2 lru_max = 2 , lru_min=1
This means that when any lru is 2% dirty flush it clean to 1%. At
checkpoint time all of the dirty buffers, which in this case will average
0.5% of all buffers (or 5,000 buffers) will have to be flushed making the
checkpoint rather short.

Another upshot of leaving 50% of your buffers dirty is the possibility that
at peak load times your applications will not be able to find a clean buffer
into which to place a new page from disk or a completely new page and will
have to perform what's known as a Foreground Write. This means that the
engine thread running your query will have to stop to write out a page on
its own. This is very bad and will severely reduce server performance. If
onstat -F shows more than a VERY small number (some say only zero is
acceptable) of foreground writes either your buffer cache is way too small
or the lru flush parameters are too high.

Art

On Tue, Feb 17, 2009 at 10:48 PM, jak aka <jakkritakeng@yahoo.com> wrote:

> Thank you all.
>
> I have few more question.
>
> Suppose.
> buffer = 1000000
>
> case1 lru_max = 51 , lru_min=50
> case2 lru_max = 2 , lru_min=1
>
> Dirty pages will be flushed 1% when checkpoint time (1% of 1M = 10000
> pages) .
> The checkpoint duration must be equal becase both case will flush only
> 10000
> pages. Am I correct ?
>
> - If I want to reduce the checkpoint to the mimimum duration , Can I set
> 0.0002, 0.0001 (lru_max,lru_min) (i'm not sure how many digit can I set) or
> 50.0002,50.0001 (0.0001 % of buffer to be flush when normal checkpoint) to
> make all dirty page flushed by lru-flush.
> If I do so, Any concern would you recommend ?
>
> Best regards,
> Jakkrit A.
>
> ________________________________
> From: John Miller iii <miller3@us.ibm.com>
> To: ids@iiug.org
> Sent: Wednesday, February 18, 2009 4:03:31 AM
> Subject: Re: BUFFERPOOL question [14905]
>
> While this answer is very accurate for IDS releases before version 11. =
> In
> version 11
> there is a new feature called non-blocking checkpoints. In short, it i=
> s
> generally
> beneficial to have lru min and max dirty above 50% in version 11 or hig=
> her.
> I
> would suggest reading the following for more details.
>
> http://www.ibm.com/developerworks/data/library/techarticle/dm-0703lashl=
> ey/index.html
>
> John F. Miller III
> STSM, Support Architect
> miller3@us.ibm.com
> 503-578-5645
> IBM Informix Dynamic Server (IDS)
>
> ids-bounces@iiug.org wrote on 02/17/2009 09:40:46 AM:
>
> > Informix have two different forms of flush dirty page from buffer to
> disck.
> > The first one is online (not lock) and if made by the cleaners theard=
> s.
> The
> > second one if in the checkpoint proccess (locking updates).
> >
> > The form of work of the cleaners depends of the lru_min_dirty and
> > lru_max_dirty parameters. Lru_max_dirty parameter specify when the
> cleaners
> > began to work and lru_min_dirty when cleaners stop to work. By exampl=
> e,
> If
> > we have a have a configuration of 51,50 (max_dirty and min_dirty). Th=
> e
> > cleaners not began to work until the 51% of the buffers are dirty. In=
>
> your
> > second example if we have 2,1, the cleaners began to work when the 2%=
> of
> the
> > buffers are dirty. Is clear than with a configuration of 2,1 we go to=
>
> have
> > minus dirty pages when the checkpoint process start and the duration =
> of
> > checkpoints will be smaller.
> >
> > This sound fine, but you also must to have in consideration than flus=
> h
> dirty
> > page is more efficient in checkpoint process than will cleaners
> (checkpoint
> > don=B4t need to worry about locks).
> >
> > So, in sumary if you have interative operations in your database your=
>
> must
> > to have small values for lry_min_dirty and lry_max_dirty and a small
> > checkpoint interval in order to not lock interatives users. But if yo=
> ur
> > database only have batch operation don't worry about checkpoint times=
>
> > because don't lock anyone and put high values for this parameters.
> >
> > I hope this clarify you
> >
> > Juan Jorge Cruces Fern=E1ndez
> > Accelya
> >
> > ----- Original Message -----
> > From: "jak aka" <jakkritakeng@yahoo.com>
> > To: <ids@iiug.org>
> > Sent: Monday, February 16, 2009 6:27 AM
> > Subject: Re: BUFFERPOOL question [14893]
> >
> > > Sorry , I may ask you not clear. I know what's the meaning of lru_m=
> ax
> and
> > > lru_min and I 've also read many book but I still not clear. I don'=
> t
> want
> > > to
> > > waste your time if you feel uncomfortable to answer please do nothi=
> ng.
> > >
> > > Anyway , my question is if I remain the gap 1 (lru_max -
> lrun_min ,2-1 ,
> > > 52-51) . What 's the different ?
> > > What will be effect, checkpoint ,... ?
> > >
> > > Sorry again Art.
> > >
> > > ________________________________
> > > From: Art Kagel <art.kagel@gmail.com>
> > > To: ids@iiug.org
> > > Sent: Thursday, February 12, 2009 6:35:43 PM
> > > Subject: Re: BUFFERPOOL question [14857]
> > >
> > > Trick question?
> > >
> > > The LRU flush parameters, lru_max_dirty and lru_min_dirty are
> different.
> > >
> > > If you want more than that, look in the Administrator's Reference
> manual
> > > for
> > > the meaning, purpose, and setting of the lru flush parameters in th=
> e
> > > BUFFERPOOL section. You may also find more information in the Tunin=
> g
> Guide
> > > or by searching the archives of this list for lru_min_dirty,
> > > lru_max_dirty,
> > > and lru_min/max_dirty (I sometimes refer to both together that way.=
>
> Meet
> > > us
> > > half way Jak, do some research before you post a question.
> > >
> > > (*&^ but I'm cranky today.
> > >
> > > Art
> > >
> > > On Thu, Feb 12, 2009 at 2:31 AM, jak aka <jakkritakeng@yahoo.com>
> wrote:
> > >
> > >> Hi all,
> > >>
> > >> I have a basic question but it make me confuse. What's the differe=
> nt
> > >> between
> > >> BUFFERPOOL below ?
> > >>
> > >> BUFFERPOOL
> > >> size=3D2K,buffers=3D1000000,lrus=3D64,lru_min_dirty=3D1,lru_max_di=
> rty=3D2
> > >> and
> > >> BUFFERPOOL
> > >> size=3D2K,buffers=3D1000000,lrus=3D64,lru_min_dirty=3D51,lru_max_d=
> irty=3D52
> > >>
> > >> Best Regards,
> > >> Jakkrit A.
> > >>
> > >>
> > >>
> > >>
> > >
> > >
> >
> ***********************************************************************=
> ********
>
> > >> Forum Note: Use "Reply" to post a response in the discussion forum=
> ...
> > >>
> > >>
> > >
> > > --
> > > Art S. Kagel
> > > Oninit (www.oninit.com)
> > > IIUG Board of Directors (art@iiug.org)
> > >
> > > Disclaimer: Please keep in mind that my own opinions are my own
> opinions
> > > and
> > > do not reflect on my employer, Oninit, the IIUG, nor any other
> > > organization
> > > with which I am associated either explicitly or implicitly. Neither=
> do
> > > those opinions reflect those of other individuals affiliated with a=
> ny
> > > entity
> > > with which I am affiliated nor those of the entities themselves.
> > >
> > > --001485f91dd2b220b50462b71e68
> > >
> > >
> > >
> >
> ***********************************************************************=
> ********
>
> > > Forum Note: Use "Reply" to post a response in the discussion forum.=
>
> > >
> > >
> > >
> >
> ***********************************************************************=
> ********
>
> > > Forum Note: Use "Reply" to post a response in the discussion forum.=
>
> > >
> > >
> >
> > Este mensaje se dirige exclusivamente a su destinatario, contiene
> informaci=F3n
> > CONFIDENCIAL sometida a secreto profesional y su divulgaci=F3n est=E1=
>
> prohibida
> > por ley. Si ha recibido este mensaje por error, su copia y uso est=E1=
> n
> > prohibidos, rog=E1ndole que nos lo comunique inmediatamente por esta
> > misma v=EDa y
> > proceda a su destrucci=F3n.
> > El correo electr=F3nico no garantiza la confidencialidad de los mensa=
> jes ni
> su
> > integridad o correcta recepci=F3n. La compa=F1=EDa ACCELYA no asume
> responsabilidad
> > por estas circunstancias.
> > Si el destinatario de este mensaje no consintiera la utilizaci=F3n de=
> l
> correo
> > electr=F3nico y la grabaci=F3n de los mensajes, rogamos nos lo notifi=
> quede
> forma
> > inmediata.
> > This message is intended exclusively for its addressee. It contains
> > information that is CONFIDENTIAL and protected by a professional
> privilege or
> > whose disclosure is prohibited by law. If this message has been recei=
> ved
> in
> > error, you should know that it is forbidden to copy or use it. Please=
>
> > immediately notify us via e-mail and delete it.
> > Internet e-mail neither guarantees the confidentiality nor the integr=
> ity
> or
> > proper receipt of the messages sent. ACCELYA company assumes no liabi=
> lity
> or
> > responsibility for any error caused by matters beyond our control. If=
> the
>
> > addressee of this message does not consent to the use of Internet e-m=
> ail
> and
> > message recording, please notify us immediately.
> >
> >
> >
> ***********************************************************************=
> ********
>
> > Forum Note: Use "Reply" to post a response in the discussion forum.=
>
> >=
>
>
>
> *******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>
>
> *******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>

--
Art S. Kagel
Oninit (www.oninit.com)
IIUG Board of Directors (art@iiug.org)

Disclaimer: Please keep in mind that my own opinions are my own opinions and
do not reflect on my employer, Oninit, the IIUG, nor any other organization
with which I am associated either explicitly or implicitly. Neither do
those opinions reflect those of other individuals affiliated with any entity
with which I am affiliated nor those of the entities themselves.

--0016364ee2ea4dc7ac0463298b99

Messages In This Thread

• BUFFERPOOL question
  jak aka -- Thursday, 12 February 2009, at 2:31 a.m.
  • AW: BUFFERPOOL question
    Andreas.KUTSCHE@spar.at -- Thursday, 12 February 2009, at 6:22 a.m.
  • Re: BUFFERPOOL question
    Art Kagel -- Thursday, 12 February 2009, at 6:35 a.m.
    • Re: BUFFERPOOL question
      jak aka -- Monday, 16 February 2009, at 12:27 a.m.
      • Re: BUFFERPOOL question
        Jorge cruces -- Tuesday, 17 February 2009, at 12:40 p.m.
        • Re: BUFFERPOOL question
          John Miller iii -- Tuesday, 17 February 2009, at 4:03 p.m.
          • Re: BUFFERPOOL question
            jak aka -- Tuesday, 17 February 2009, at 10:48 p.m.
            • Re: BUFFERPOOL question
              Art Kagel -- Tuesday, 17 February 2009, at 11:06 p.m.
              • Re: BUFFERPOOL question
                jak aka -- Wednesday, 18 February 2009, at 11:05 p.m.
      • Re: BUFFERPOOL question
        Art Kagel -- Tuesday, 17 February 2009, at 7:33 p.m.
  • Re: BUFFERPOOL question
    Mark Jamison -- Friday, 13 February 2009, at 2:18 p.m.

IDS Forum is maintained by Administrator with WebBBS 5.12.