I/O performance problem after change to HP-UX 11.31
After migrating IDS 10.00.FC8 from HP-UX/PA-RISC to Itanium (with a new EMC SAN cabinet and switches), sequential scans ran about three times slower despite identical IDS config, low CPU and good cache rates. Suggestions included benchmarking raw devices with dd to see whether Informix was implicated, checking SAN RAID level and stripe sizes, and enabling KAIO. The dd tests confirmed the new storage itself was slower: the databases had landed on a slower class of disk in the EMC array. Moving them to the faster disks restored the original dd and IDS I/O performance.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
We are using 10.00.FC8 on HP-UX Moved all our databases from Parisc to Itanium this weekend and are experiencing that disk io is much slower. We have exactly the same IDS config, same database layout in both cases. Data is stored in a SAN (EMC), the new production environment has a new EMC cabinet, new SAN switches. As an example a seq scan of a large table takes about 3 times the time it took in the old environment. CPU load is not noticable, cache rate is fine. We are not using KAIO. Any ideas that can explain this ? TIA Ulf
"Ulf" <ulf.akerberg@gmail.com> wrote in message news:2f85e555-2db8-4456-8b8c-f63ea068b744@k2g2000yql.googlegroups.com... > > > We are using 10.00.FC8 on HP-UX > > Moved all our databases from Parisc to Itanium this weekend and are > experiencing that disk io is much slower. > > We have exactly the same IDS config, same database layout in both > cases. Data is stored in a SAN (EMC), the new production environment > has a new EMC cabinet, new SAN switches. > > As an example a seq scan of a large table takes about 3 times the > time it took in the old environment. CPU load is not noticable, cache > rate is fine. > > We are not using KAIO. > > Any ideas that can explain this ? Are you able to eliminate, or implicate, Informix as the cause by doing some dd's directly from the RISC and Itanium servers? eg time dd if=/dev/zero of=/dev/vg01/rtest1 bs=2k count=1000000 Can you post the results pls?
How is the new SAN configured? HP and EMC are fond of RAID5 which will definitely slow you down. EMC is also fond of very wide stripe block sizes which will also slow you down. Why not use KAIO? It is also faster! 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. On Mon, May 25, 2009 at 3:04 PM, Ulf <ulf.akerberg@gmail.com> wrote: > > > We are using 10.00.FC8 on HP-UX > > Moved all our databases from Parisc to Itanium this weekend and are > experiencing that disk io is much slower. > > We have exactly the same IDS config, same database layout in both > cases. Data is stored in a SAN (EMC), the new production environment > has a new EMC cabinet, new SAN switches. > > As an example a seq scan of a large table takes about 3 times the > time it took in the old environment. CPU load is not noticable, cache > rate is fine. > > We are not using KAIO. > > Any ideas that can explain this ? > > TIA > > Ulf > _______________________________________________ > Informix-list mailing list > Informix-list@iiug.org > http://www.iiug.org/mailman/listinfo/informix-list >
Ulf wrote:
>
> We are using 10.00.FC8 on HP-UX
>
> Moved all our databases from Parisc to Itanium this weekend and are
> experiencing that disk io is much slower.
>
> We have exactly the same IDS config, same database layout in both
> cases. Data is stored in a SAN (EMC), the new production environment
> has a new EMC cabinet, new SAN switches.
>
> As an example a seq scan of a large table takes about 3 times the
> time it took in the old environment. CPU load is not noticable, cache
> rate is fine.
>
> We are not using KAIO.
>
> Any ideas that can explain this ?
>
> TIA
>
> Ulf
What does "file oninit" show?
On 26 Maj, 08:59, theBP <th...@Usenet-News.Net> wrote:
> Ulf wrote:
>
> > We are using 10.00.FC8 on HP-UX
>
> > Moved all our databases from Parisc to Itanium this weekend and are
> > experiencing that disk io is much slower.
>
> > We have exactly the same IDS config, same database layout in both
> > cases. Data is stored in a SAN (EMC), the new production environment
> > has a new EMC cabinet, new SAN switches.
>
> > As an example a seq scan of a large table takes about 3 times the
> > time it took in the old environment. CPU load is not noticable, cache
> > rate is fine.
>
> > We are not using KAIO.
>
> > Any ideas that can explain this ?
>
> > TIA
>
> > Ulf
>
> What does "file oninit" show?- Dölj citerad text -
>
> - Visa citerad text -
file oninit show: oninit: ELF-64 executable object file - IA64
On May 25, 9:04 pm, Ulf <ulf.akerb...@gmail.com> wrote: > We are using 10.00.FC8 on HP-UX > > Moved all our databases from Parisc to Itanium this weekend and are > experiencing that disk io is much slower. > > We have exactly the same IDS config, same database layout in both > cases. Data is stored in a SAN (EMC), the new production environment > has a new EMC cabinet, new SAN switches. > > As an example a seq scan of a large table takes about 3 times the > time it took in the old environment. CPU load is not noticable, cache > rate is fine. > > We are not using KAIO. > > Any ideas that can explain this ? > > TIA > > Ulf HP says it will not port HP-UX to X86 but HP alone will not be able maintain Itanium "momentum" for ever. So I would thing the actual choices are SUN (Sparc and X86 Solaris) or IBM Power. I would like to understand why companies decide still to migrate to Itanium and how they motivate technical people to solve problems on that particular platform (knowing very well there is no fully problem-free platform, but there are platforms that do disappear). Bernard Dhooghe
On 25 Maj, 23:45, "Neil Truby" <neil.tr...@ardenta.com> wrote: > "Ulf" <ulf.akerb...@gmail.com> wrote in message > > news:2f85e555-2db8-4456-8b8c-f63ea068b744@k2g2000yql.googlegroups.com... > > > > > > > > > We are using 10.00.FC8 on HP-UX > > > Moved all our databases from Parisc to Itanium this weekend and are > > experiencing that disk io is much slower. > > > We have exactly the same IDS config, same database layout in both > > cases. Data is stored in a SAN (EMC), the new production environment > > has a new EMC cabinet, new SAN switches. > > > As an example a seq scan of a large table takes about 3 times the > > time it took in the old environment. CPU load is not noticable, cache > > rate is fine. > > > We are not using KAIO. > > > Any ideas that can explain this ? > > Are you able to eliminate, or implicate, Informix as the cause by doing some > dd's directly from the RISC and Itanium servers? > eg > > time dd if=/dev/zero of=/dev/vg01/rtest1 bs=2k count=1000000 > > Can you post the results pls?- Dölj citerad text - > > - Visa citerad text - Tests using dd showed that the new environment was much slower. This was due to the fact that the disks used in the EMC cabinet was slower that the ones used before. We have now moved our databases to the other, faster type, of disks. The result is that we get the same speed using dd as before and the same IO performance from IDS as before. Thanks for your input ! Ulf