Re: Advise needed - BUFFERS increased?
Posted in 2003
Topics: Performance & Tuning, Logging & Checkpoints
Everything looks good actually. The BTR of 8.1 is just a smidge high but not
worrysome so I checked the onstat -P output. You have over 800 unused buffers
or about 1.3% which is fine and along with the low BR and acceptable BTR
indicates that you have enough buffers. Your seqscans is very high but likely
are being performed on smaller lookup tables since the BTR does not seem to be
churning. All in all a well tuned system.
Certainly adding buffers will not hurt the IDS server at all, but if you are
experiencing swapping already it will exacerbate that situation. Perhaps you
can move non-server apps off onto another machine and make this a dedicated IDS
server? With a fast network there will be no additional latency and freeing
resources on the app and server machines will improve throughput and
responsiveness.
BTW what is the 'problem' you are looking to solve? You never mentioned it!
Art S. Kagel
----- Original Message -----
From: Milena Ilieva <milena_ilieva@yahoo.com>
At: 2/ 6 15:38
> Hello,
>
> I need to hear some other opinion about the problem we
> are having with the Informix performance and the
> change I am planning to do.
>
> I have calculated the BR (buffer ratio) and it looks
> not so bad =7.2%, the BTR = 8.1.
> I had increased the LRUS to 120 a while ago and it the
> BR ususaly is fine since. The question was about the
> number of BUFFERS=60000 - whether it is a bottlneck or
> not. My only concern is that the Read cache is > 95%
> and the checkpoints are 0-1 sec. According to Informix
> recommendation - BUFFERS should take about 20-25% of
> the physical memory. We have 1.5 Gb physical memory,
> 20% from it = 175 000 Kb. So we are quite under what
> is recommended.
> My question is in case the checkpoints are OK and the
> Read cache is OK as well, should I increase the number
> of the BUFFERS anyway?
>
> I noticed that there si some paging on the Unix side
> as well and some of the values exceed a lot of what
> Caldera recommends, vflt/s should be up to 50 and in
> our case sometimes (usualy when there si a slowdown)
> it is very high. We have opened a case with Caldera
> and are waiting for them get back to us if there si
> any Unix kernel parameters tuned.
> Here are some outputs:
> informix@vaprp1> sar -p 5 5
>
> UnixWare vaprp1 5 7.1.1 i386 02/06/03
>
> 10:18:57 atch/s atfree/s atmiss/s pgin/s ppgin/s
> pflt/s vflt/s slock/s
> 10:19:02 577.26 46.59 19.90 12.54 26.09
> 161.38 3237.51 0.00
> 10:19:07 642.43 41.83 16.73 13.15 13.15
> 118.13 1160.16 0.00
> 10:19:12 415.34 33.07 15.74 7.17 8.37
> 121.12 2102.99 0.00
> 10:19:17 489.64 79.88 41.24 28.88 28.88
> 184.26 3093.03 0.00
> 10:19:22 279.28 33.86 16.53 13.15 13.15
> 76.29 2037.45 0.00
> Average 480.76 47.04 22.03 14.98 17.92
> 132.23 2325.75 0.00
>
>
> informix@vaprp1> sar -g 5 5
>
> UnixWare vaprp1 5 7.1.1 i386 02/06/03
>
> 10:26:01 pgout/s ppgout/s vfree/s pfree/s
> vscan/s
> 10:26:06 0.00 0.00 0.00 0.00
> 0.00
> 10:26:11 0.00 0.00 0.00 0.00
> 0.00
> 10:26:16 0.00 0.00 0.00 0.00
> 0.00
> 10:26:21 0.00 0.00 0.00 0.00
> 0.00
> 10:26:26 0.00 0.00 0.00 0.00
> 0.00
> Average 0.00 0.00 0.00 0.00
> 0.00
>
>
> informix@vaprp1> onstat -p
>
> Informix Dynamic Server Version 7.30.UC10 -- On-Line
> -- Up 10 days 00:22:11 --
> 385408 Kbytes
>
> Profile
> dskreads pagreads bufreads ched dskwrits pagwrits
> bufwrits ched
> 44696909 10248696 1167888021 96.17 7258168 6603468
> 107632833 93.26
>
> isamtot open start read write rewrite
> delete commit rollbk
> 3863981449 59800856 251343347 2931610071 33553037
> 5332563 8340525 2942299 61
>
> gp_read gp_write gp_rewrt gp_del gp_alloc gp_free
> gp_curs
> 0 0 0 0 0 0
> 0
>
> ovlock ovuserthread ovbuff usercpu syscpu
> numckpts flushes
> 0 0 0 231653.57 15413.57 2994
> 5988
>
> bufwaits lokwaits lockreqs deadlks dltouts ckpwaits
> compress seqscans
> 8494035 7352 1324200902 0 0 6624
> 891354 14479158
>
> ixda-RA idx-RA da-RA RA-pgsused lchwaits
> 13176595 182578 11571397 24914619 35550
>
>
> Message Log File: /usr2/informix/online.log
> 09:21:55 Checkpoint Completed: duration was 0
> seconds.
> 09:26:55 Checkpoint Completed: duration was 0
> seconds.
> 09:31:55 Checkpoint Completed: duration was 0
> seconds.
> 09:36:55 Checkpoint Completed: duration was 0
> seconds.
> 09:38:09 Logical Log 192586 Complete.
> 09:38:11 Logical Log 192586 - Backup Started
> 09:38:30 Logical Log 192586 - Backup Completed
> 09:41:55 Checkpoint Completed: duration was 0
> seconds.
> 09:46:55 Checkpoint Completed: duration was 0
> seconds.
> 09:51:55 Checkpoint Completed: duration was 0
> seconds.
> 09:56:55 Checkpoint Completed: duration was 0
> seconds.
> 10:01:56 Checkpoint Completed: duration was 0
> seconds.
> 10:06:56 Checkpoint Completed: duration was 0
> seconds.
> 10:11:48 Logical Log 192587 Complete.
> 10:11:51 Logical Log 192587 - Backup Started
> 10:11:56 Checkpoint Completed: duration was 1
> seconds.
> 10:12:01 Logical Log 192587 - Backup Completed
> 10:16:57 Checkpoint Completed: duration was 1
> seconds.
> 10:21:58 Checkpoint Completed: duration was 0
> seconds.
> 10:26:58 Checkpoint Completed: duration was 1
> seconds.
>
> informix@vaprp1> onstat -P|more
>
> Informix Dynamic Server Version 7.30.UC10 -- On-Line
> -- Up 10 days 00:25:58 --
> 385408 Kbytes
> partnum total btree data other resident
> dirty
> 0 1283 270 130 883 0
> 227
> 1048578 2 1 1 0 0
> 0
> 1048579 2 2 0 0 0
> 0
> 1048580 1 1 0 0 0
> 0
> 1048582 1 1 0 0 0
> 0
> 1048584 1 1 0 0 0
> 0
> 1048606 1 1 0 0 0
> 0
> 1048708 2 1 0 1 0
> 0
> 1048709 2 0 1 1 0
> 0
> 3145743 1 0 1 0 0
> 1
> 3145746 2 0 1 1 0
> 0
> 3145779 2 2 0 0 0
> 0
> 3145786 1 0 1 0 0
> 1
> 3145803 1 0 1 0 0
> 0
> 3145835 1 1 0 0 0
> 0
> 3145862 2 2 0 0 0
> 0
> 3145883 1 1 0 0 0
> 0
> 3145923 5 0 5 0 0
> 0
> 3145985 2 0 2 0 0
> 2
>
> Thanks,
>
> Milena
>
>
>
> __________________________________________________
> Do you Yahoo!?
> Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
> http://mailplus.yahoo.com
It is a
performance problem.
I just increased the number of the BUFFERS to 150000.
I will be monitoring the system and if it gets worst I
will lower the number.
I am looking at the ouput of sar -b:
informix@vaprp1> sar -b 5 5
UnixWare vaprp1 5 7.1.1 i386 02/06/03
13:08:29 bread/s lread/s %rcache bwrit/s lwrit/s
%wcache pread/s pwrite/s
13:08:34 0 80 100 14 16
13 0 0
13:08:39 0 146 100 27 36
25 0 0
13:08:44 0 120 100 22 22
3 0 0
13:08:49 0 206 100 36 38
5 0 0
13:08:54 0 171 100 25 30
16 0 0
Average 0 145 100 25 28
13 0 0
The write cache is quite low - the recommended value
is 60%. I am not sure if Unix tuning will help us with
the performance problems?
We have three applications and all three of them are
Informix applications. We are thinking to move one of
them on a separate machine, since it works with
separate DB, not the production one. But before going
ahead with that I want to make sure we have done
everything so that the performance to go away.Thanks,
Milena
--- "ART KAGEL, ...." <KAGEL@bloomberg.net> wrote:
>
> Everything looks good actually. The BTR of 8.1 is
> just a smidge high but not
> worrysome so I checked the onstat -P output. You
> have over 800 unused buffers
> or about 1.3% which is fine and along with the low
> BR and acceptable BTR
> indicates that you have enough buffers. Your
> seqscans is very high but likely
> are being performed on smaller lookup tables since
> the BTR does not seem to be
> churning. All in all a well tuned system.
>
> Certainly adding buffers will not hurt the IDS
> server at all, but if you are
> experiencing swapping already it will exacerbate
> that situation. Perhaps you
> can move non-server apps off onto another machine
> and make this a dedicated IDS
> server? With a fast network there will be no
> additional latency and freeing
> resources on the app and server machines will
> improve throughput and
> responsiveness.
>
> BTW what is the 'problem' you are looking to solve?
> You never mentioned it!
>
> Art S. Kagel
>
> ----- Original Message -----
> From: Milena Ilieva <milena_ilieva@yahoo.com>
> At: 2/ 6 15:38
>
> > Hello,
> >
> > I need to hear some other opinion about the
> problem we
> > are having with the Informix performance and the
> > change I am planning to do.
> >
> > I have calculated the BR (buffer ratio) and it
> looks
> > not so bad =7.2%, the BTR = 8.1.
> > I had increased the LRUS to 120 a while ago and it
> the
> > BR ususaly is fine since. The question was about
> the
> > number of BUFFERS=60000 - whether it is a
> bottlneck or
> > not. My only concern is that the Read cache is >
> 95%
> > and the checkpoints are 0-1 sec. According to
> Informix
> > recommendation - BUFFERS should take about 20-25%
> of
> > the physical memory. We have 1.5 Gb physical
> memory,
> > 20% from it = 175 000 Kb. So we are quite under
> what
> > is recommended.
> > My question is in case the checkpoints are OK and
> the
> > Read cache is OK as well, should I increase the
> number
> > of the BUFFERS anyway?
> >
> > I noticed that there si some paging on the Unix
> side
> > as well and some of the values exceed a lot of
> what
> > Caldera recommends, vflt/s should be up to 50 and
> in
> > our case sometimes (usualy when there si a
> slowdown)
> > it is very high. We have opened a case with
> Caldera
> > and are waiting for them get back to us if there
> si
> > any Unix kernel parameters tuned.
> > Here are some outputs:
> > informix@vaprp1> sar -p 5 5
> >
> > UnixWare vaprp1 5 7.1.1 i386 02/06/03
> >
> > 10:18:57 atch/s atfree/s atmiss/s pgin/s
> ppgin/s
> > pflt/s vflt/s slock/s
> > 10:19:02 577.26 46.59 19.90 12.54
> 26.09
> > 161.38 3237.51 0.00
> > 10:19:07 642.43 41.83 16.73 13.15
> 13.15
> > 118.13 1160.16 0.00
> > 10:19:12 415.34 33.07 15.74 7.17
> 8.37
> > 121.12 2102.99 0.00
> > 10:19:17 489.64 79.88 41.24 28.88
> 28.88
> > 184.26 3093.03 0.00
> > 10:19:22 279.28 33.86 16.53 13.15
> 13.15
> > 76.29 2037.45 0.00
> > Average 480.76 47.04 22.03 14.98
> 17.92
> > 132.23 2325.75 0.00
> >
> >
> > informix@vaprp1> sar -g 5 5
> >
> > UnixWare vaprp1 5 7.1.1 i386 02/06/03
> >
> > 10:26:01 pgout/s ppgout/s vfree/s pfree/s
> > vscan/s
> > 10:26:06 0.00 0.00 0.00 0.00
> > 0.00
> > 10:26:11 0.00 0.00 0.00 0.00
> > 0.00
> > 10:26:16 0.00 0.00 0.00 0.00
> > 0.00
> > 10:26:21 0.00 0.00 0.00 0.00
> > 0.00
> > 10:26:26 0.00 0.00 0.00 0.00
> > 0.00
> > Average 0.00 0.00 0.00 0.00
> > 0.00
> >
> >
> > informix@vaprp1> onstat -p
> >
> > Informix Dynamic Server Version 7.30.UC10 --
> On-Line
> > -- Up 10 days 00:22:11 --
> > 385408 Kbytes
> >
> > Profile
> > dskreads pagreads bufreads ched dskwrits pagwrits
> > bufwrits ched
> > 44696909 10248696 1167888021 96.17 7258168
> 6603468
> > 107632833 93.26
> >
> > isamtot open start read write
> rewrite
> > delete commit rollbk
> > 3863981449 59800856 251343347 2931610071 33553037
> > 5332563 8340525 2942299 61
> >
> > gp_read gp_write gp_rewrt gp_del gp_alloc
> gp_free
> > gp_curs
> > 0 0 0 0 0 0
> > 0
> >
> > ovlock ovuserthread ovbuff usercpu syscpu
> > numckpts flushes
> > 0 0 0 231653.57 15413.57
> 2994
> > 5988
> >
> > bufwaits lokwaits lockreqs deadlks dltouts
> ckpwaits
> > compress seqscans
> > 8494035 7352 1324200902 0 0
> 6624
> > 891354 14479158
> >
> > ixda-RA idx-RA da-RA RA-pgsused lchwaits
> > 13176595 182578 11571397 24914619 35550
> >
> >
> > Message Log File: /usr2/informix/online.log
> > 09:21:55 Checkpoint Completed: duration was 0
> > seconds.
> > 09:26:55 Checkpoint Completed: duration was 0
> > seconds.
> > 09:31:55 Checkpoint Completed: duration was 0
> > seconds.
> > 09:36:55 Checkpoint Completed: duration was 0
> > seconds.
> > 09:38:09 Logical Log 192586 Complete.
> > 09:38:11 Logical Log 192586 - Backup Started
> > 09:38:30 Logical Log 192586 - Backup Completed
> > 09:41:55 Checkpoint Completed: duration was 0
> > seconds.
> > 09:46:55 Checkpoint Completed: duration was 0
> > seconds.
> > 09:51:55 Checkpoint Completed: duration was 0
> > seconds.
> > 09:56:55 Checkpoint Completed: duration was 0
> > seconds.
> > 10:01:56 Checkpoint Completed: duration was 0
> > seconds.
> > 10:06:56 Checkpoint Completed: duration was 0
> > seconds.
> > 10:11:48 Logical Log 192587 Complete.
> > 10:11:51 Logical Log 192587 - Backup Started
> > 10:11:56 Checkpoint Completed: duration was 1
> > seconds.
> > 10:12:01 Logical Log 192587 - Backup Completed
> > 10:16:57 Checkpoint Completed: duration was 1
>
=== message truncated ===
__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com