4GL extremely slowed down after migration
Posted in 2004
Topics: Connectivity: ESQL/C, 4GL & Embedded SQL, Migration, Import/Export & Data Conversion, Internationalization & Character Sets, Versions, Editions & End-of-Life
We have
migrated (dbexport/dbimport) database from:
2 x 800 MHz P3, 512 MB RAM, 3 logical disks (every one is RAID-1 made of two
18.2 GB 10K RPM disks, using Compaq RAID controller, can't remember exact
designation), SCO OpenServer 5.0.5, IDS 7.30.UC2. BTW, used only 1 CPU for
IDS. The CPU consumption was moderate (below 40% mostly).
to:
1 x 2.2 GHz Xeon (P4), 2 GB RAM, 6 logical disks (every one is RAID-1 made of
two disks) on 2 Compaq 641 RAID controllers. SCO OpenServer 5.0.6 (with
recommended patches), IDS 7.31.UD4.
On RAID controllers I always turn off write cache.
4GL (ver. 6.0.2) programs are about 10 times slower on new machine. I noticed
that CPU is >90% used, and disks are mostly idle.
I did update statistics in various ways, even for the whole database (HIGH).
The database size iz about 13-14 GB, but 80% of the size is in few large
tables that are not accessed by those 4GL programs. Programmers did recompile
4GLs on the new machine, and we also tried to use old binaries.
Is there any known bug in such a combination of rather fresh HW and engine and
ancient OS and 4GL? Any other idea? We need it very urgently.
TIA,
Darko
Dear Mr Krstic,
It is not possible to determine the problem with the information you send.
What processes are using the CPU ?
I guess it's the oninit(s) .
What is the oninit doing? You may start with 'onstat -g act' to monitor the
system for a
while.
What are the Informix threads are wainting for (mainly) ? If you have
enabled
wait staistics (WSTATS in $ONCONFIG) you may use the follwing SQL query to
obtain
a picture:
database sysmaster;select sum(cumtime), reason from sysmaster:sysseswts
where cumtime >1000 and reason != 'condition'
group by reason
order by 1 desc;
Is there anything peculiar in onstat -g ath (many theads waiting for a
resource) ?
UPSTAT HIGH for the whole database is probably an overkill. Usually it will
be enough
to run high / medium for all indexed columns (provided all the relevant
clomns have been
indexed) .
With kind regards
Tilman Model-Bosch
--
Tilman Model-Bosch
IBM Data Managment Solutions, Informix Advanced Support
c\\\\o SAP AG
TECHDEV 05
Neurrotstr.16
69190 Walldorf
forum.subscriber@iiug.org wrote on 18/07/2004 22:40:22:
> We have migrated (dbexport/dbimport) database from:
> 2 x 800 MHz P3, 512 MB RAM, 3 logical disks (every one is RAID-1
> made of two 18.2 GB 10K RPM disks, using Compaq RAID controller,
> can't remember exact designation), SCO OpenServer 5.0.5, IDS 7.30.
> UC2. BTW, used only 1 CPU for IDS. The CPU consumption was moderate
> (below 40% mostly).
> to:
> 1 x 2.2 GHz Xeon (P4), 2 GB RAM, 6 logical disks (every one is
> RAID-1 made of two disks) on 2 Compaq 641 RAID controllers. SCO
> OpenServer 5.0.6 (with recommended patches), IDS 7.31.UD4.
>
> On RAID controllers I always turn off write cache.
>
> 4GL (ver. 6.0.2) programs are about 10 times slower on new machine.
> I noticed that CPU is >90% used, and disks are mostly idle.
>
> I did update statistics in various ways, even for the whole database
> (HIGH). The database size iz about 13-14 GB, but 80% of the size is
> in few large tables that are not accessed by those 4GL programs.
> Programmers did recompile 4GLs on the new machine, and we also tried
> to use old binaries.
>
> Is there any known bug in such a combination of rather fresh HW and
> engine and ancient OS and 4GL? Any other idea? We need it very urgently.
>
> TIA,
> Darko
>
Related threads
- Posting from the Informix-list
- Migrating from IDS 9.40.UC6 to 11.50.UC3
- Ip for a network session
- questions onstat -g