Big Performance differences between (almost) identical servers
Hi, there.
I'm dealing with an IDS 11.5 FC6 productive dataserver under Solaris 9 with an V890 a little old - almost five year running - with a lot of performance issues running SPs. For instance, there are a SP which takes 37 second to do the work. Looking at code we haven't found nothing specially wrong - in fact when we run each query into SP using dbaccess and replacing parameters manually each one runs ok -. Tables used by SP are fine, with all necessary indexes, update statistics and onchecks in order, and views used also are well-declared (actually, some sps with problems uses several views ). This problem doesn't affect other sps, only some of them which are used a lot. I forgot to say: we get samei behaviour with a lot, few or zero users, with server up since 2 days, a week or just after restarting instance.
By the other side, I have access to a backup server: IDS 11.5 FC6 under solaris 9, also a V890 but a "younger" one - less than a year, one of the last of that model -. We copy informix data restoring ontapes level-0 we do on the productive one. when we run same test we did on productive server everything runs ok: our SP for tests runs in 0-1 seconnd instead of 37, other affected sp runs fine also. I used same onconfig file in order to get identical running environment.
Differences between servers - as far as I can see - are that backup server hardware are newer, with less processors (!!), and uses storage manager (Legatto I think) to manage data areas, and productive server is older - but not too much - more processors and uses raw areas for storing data. Since we copied data to backup server using ontape data they are identical in every aspect, so it's not an index-related or data issue. It's not a configuration problem too, because we used same onconfig file in both instances, Am I wrong?
I'm running out of ideas, and before proceeding to an ontape restauration in productive server against itself - assuming that full restore are the other big difference and it could "fix" problem by miracle whatever it was (no reason to support that idea, but I'm desperate) - or tell my boss we should put our productive server in trash can and get another - It won't be a very popular idea - I ask you for any ideas, comments or advices to enlighten me.
Thanks in advance
Omar Muñoz
\
Hi:
Archiving took 1 hr. and 15 min
Restoring took 53 min.
thanks
----- Original Message ----
From: Neil Truby <neil.truby@ardenta.com>
To: informix-list@iiug.org
Sent: Sat, June 26, 2010 5:19:07 PM
Subject: Re: Big Performance differences between (almost) identical servers
>From the online log: what are the times for:
- archiving the database
- restoring it on the new kit?
"Omar Muñoz" <omarmun@yahoo.com> wrote in message
news:mailman.308.1277590509.1071.informix-list@iiug.org...
Hi, there.
I'm dealing with an IDS 11.5 FC6 productive dataserver under Solaris 9 with
an V890 a little old - almost five year running - with a lot of performance
issues running SPs. For instance, there are a SP which takes 37 second to do
the work. Looking at code we haven't found nothing specially wrong - in fact
when we run each query into SP using dbaccess and replacing parameters
manually each one runs ok -. Tables used by SP are fine, with all necessary
indexes, update statistics and onchecks in order, and views used also are
well-declared (actually, some sps with problems uses several views ). This
problem doesn't affect other sps, only some of them which are used a lot. I
forgot to say: we get samei behaviour with a lot, few or zero users, with
server up since 2 days, a week or just after restarting instance.
By the other side, I have access to a backup server: IDS 11.5 FC6 under
solaris 9, also a V890 but a "younger" one - less than a year, one of the
last of that model -. We copy informix data restoring ontapes level-0 we do
on the productive one. when we run same test we did on productive server
everything runs ok: our SP for tests runs in 0-1 seconnd instead of 37,
other affected sp runs fine also. I used same onconfig file in order to get
identical running environment.
Differences between servers - as far as I can see - are that backup server
hardware are newer, with less processors (!!), and uses storage manager
(Legatto I think) to manage data areas, and productive server is older - but
not too much - more processors and uses raw areas for storing data. Since we
copied data to backup server using ontape data they are identical in every
aspect, so it's not an index-related or data issue. It's not a configuration
problem too, because we used same onconfig file in both instances, Am I
wrong?
I'm running out of ideas, and before proceeding to an ontape restauration in
productive server against itself - assuming that full restore are the other
big difference and it could "fix" problem by miracle whatever it was (no
reason to support that idea, but I'm desperate) - or tell my boss we should
put our productive server in trash can and get another - It won't be a very
popular idea - I ask you for any ideas, comments or advices to enlighten me.
Thanks in advance
Omar Muñoz
_______________________________________________
Informix-list mailing list
Informix-list@iiug.org
http://www.iiug.org/mailman/listinfo/informix-list
Hmm. Wondered if it might be disk ... Not enough variance there probably.
Are you able to do some dd tests with 2k block sizes on each site?
Are you able t o run a single, repeatable test on a quiescent system and
check onstat -p, -D , -g iov and if possible sar -d stats are for the
respective servers after an identical test?
I'm guessing that one has Sparc III or IV procs, and the other VI+?
"Omar Mu'oz" <omarmun@yahoo.com> wrote in message
news:mailman.309.1277591678.1071.informix-list@iiug.org...
Hi
Archiving took 1 hr. and 15 min
Restoring took 53 min.
thanks
----- Original Message ----
From: Neil Truby <neil.truby@ardenta.com>
To: informix-list@iiug.org
Sent: Sat, June 26, 2010 5:19:07 PM
Subject: Re: Big Performance differences between (almost) identical servers
>From the online log: what are the times for:
- archiving the database
- restoring it on the new kit?
"Omar Mu'oz" <omarmun@yahoo.com> wrote in message
news:mailman.308.1277590509.1071.informix-list@iiug.org...
Hi, there.
I'm dealing with an IDS 11.5 FC6 productive dataserver under Solaris 9 with
an V890 a little old - almost five year running - with a lot of performance
issues running SPs. For instance, there are a SP which takes 37 second to do
the work. Looking at code we haven't found nothing specially wrong - in fact
when we run each query into SP using dbaccess and replacing parameters
manually each one runs ok -. Tables used by SP are fine, with all necessary
indexes, update statistics and onchecks in order, and views used also are
well-declared (actually, some sps with problems uses several views ). This
problem doesn't affect other sps, only some of them which are used a lot. I
forgot to say: we get samei behaviour with a lot, few or zero users, with
server up since 2 days, a week or just after restarting instance.
By the other side, I have access to a backup server: IDS 11.5 FC6 under
solaris 9, also a V890 but a "younger" one - less than a year, one of the
last of that model -. We copy informix data restoring ontapes level-0 we do
on the productive one. when we run same test we did on productive server
everything runs ok: our SP for tests runs in 0-1 seconnd instead of 37,
other affected sp runs fine also. I used same onconfig file in order to get
identical running environment.
Differences between servers - as far as I can see - are that backup server
hardware are newer, with less processors (!!), and uses storage manager
(Legatto I think) to manage data areas, and productive server is older - but
not too much - more processors and uses raw areas for storing data. Since we
copied data to backup server using ontape data they are identical in every
aspect, so it's not an index-related or data issue. It's not a configuration
problem too, because we used same onconfig file in both instances, Am I
wrong?
I'm running out of ideas, and before proceeding to an ontape restauration in
productive server against itself - assuming that full restore are the other
big difference and it could "fix" problem by miracle whatever it was (no
reason to support that idea, but I'm desperate) - or tell my boss we should
put our productive server in trash can and get another - It won't be a very
popular idea - I ask you for any ideas, comments or advices to enlighten me.
Thanks in advance
Omar Mu'oz
_______________________________________________
Informix-list mailing list
Informix-list@iiug.org
http://www.iiug.org/mailman/listinfo/informix-list