cancel
Showing results for 
Search instead for 
Did you mean: 

10gR2 RAC on HP UX IA 11.31

Naveed Ahmed_1
Occasional Visitor

10gR2 RAC on HP UX IA 11.31

Our database was 10.2.0.3 residing on Windows 2003 with 4 Intel Dual Core CPUs(1.9GHz each),6GB of RAM and hardware is IBM X series 255
Following are the characteristics of the old database
1. Database character set is WE8MSWIN1252
2. Block size is 16k
3. Datafiles on SAN
4. Standalone database
5. Archive Log database
6. SGA was 1504M
7. PGA was 194M
8. Processes parameter set to 3000
9. Open_cursors parameter set to 3000
10. Dedicated Server

We have migrated this database to a new machine using Data pump last week.
The Hardware is HP Superdome, on HP UX 11.31 Itanium with one 1.6 GHz CPU and 6GB RAM on each node
Following is the configuration of the new database
1.The database is a Two node RAC database.
2. Version 10.2.0.3
3. Archive Log enabled
4. Flash back enabled
5. DR setup with protection mode set to Max Availability
6. Block size set to 32k
7. Database storage used is ASM over Raw disks
8. Database character set is WE8MSWIN1252
9. SGA was 1654M
10. PGA was 330M
11. Processes parameter set to 3000
12. Open_cursors parameter set to 3000

The database on this new environment is performing relatively slower compared to the old environment which is quite surprising as we have a better configuration now on much better machines.
Tha application is packaged application which connects from an application server.
There is no change in the network configuration.

We have reviewed all the init.ora parameters and kernel parameters.

We want to know , has anyone migrated to HP UX 11.31 and facing performance issues and what was the workaround to tackle this.

Thanks
Naveed Ahmed
6 REPLIES
Duncan Edmonstone
Honored Contributor

Re: 10gR2 RAC on HP UX IA 11.31

Naveed,

If you'd picked up a database and moved it to HP-UX without changing anything else, I think we could maybe look at helping you here. As it is, you've changed so many other items:

Looks like you've gone from 32-bit to 64-bit
From standalone to RAC
Have you also gone from filesystems to ASM?
You've enabled flashback
You've configured Dataguard

That's a lot of changes which could effect performance so its hard to give any specific advice.

It's interesting that you state you've gone from 8 cores in total on a standalone x86, to just 2 IA64 cores spread across 2 systems in a RAC config and you think that performance should be better... I wouldn't necessarily think that!

Are you actually seeing anything in terms of CPU/memory/IO issues on the two systems?

HTH

Duncan

HTH

Duncan
Naveed Ahmed_1
Occasional Visitor

Re: 10gR2 RAC on HP UX IA 11.31

Hi Duncan,
Yes we have moved from file system to ASM, I guess this much better than a DB on filesystem.
Yes we see high memory utlization and high swapping too.

But isn't 2 IA64 cores better in terms of performance compared CPUs on Windows ?

For time being we have disabled Data guard too and but still we see the slowness in performance.

Thanks
Naveed Ahmed
Steven E. Protter
Exalted Contributor

Re: 10gR2 RAC on HP UX IA 11.31

Shalom,

Merely moving the database to the superdome will not finish the job.

Oracle needs to be tuned, the SGA will need to be adjusted, the init.ora file will require change, Oracle has a self tuning mechanism that has to be set up to collect data and make decisions.

Also, the OS probably needs performance patching. Kernel parameters probably need to be changed.

There are no short cuts.
http://h71028.www7.hp.com/erc/downloads/4aa1-0961enw.pdf

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=213172

SEP
Steven E Protter
Owner of ISN Corporation
http://isnamerica.com
http://hpuxconsulting.com
Sponsor: http://hpux.ws
Twitter: http://twitter.com/hpuxlinux
Founder http://newdatacloud.com
Naveed Ahmed_1
Occasional Visitor

Re: 10gR2 RAC on HP UX IA 11.31

We have increased the SGA size, increased the PGA size compared to what it was is in the old system but even that didnt help.

Thanks
Naveed Ahmed
Duncan Edmonstone
Honored Contributor

Re: 10gR2 RAC on HP UX IA 11.31

>> But isn't 2 IA64 cores better in terms of performance compared CPUs on Windows ?

Well that depends... I like to think of CPUs like vehicles - the Xeon is maybe a sports car, it goes fast (higher clock speed), but only carries 2 people, whilst the IA64 CPU is a bus - doesn't go as fast (lower clock speed), but can carry a lot more passengers. What works for you depends on how many 'passengers' you have (i.e. what sort of workload you have)

You state you are seeing swapping and paging... thats not good... can you post the output of:

vmstat 2 10
swapinfo -tam
kctune | grep filecache

HTH

Duncan

HTH

Duncan
Naveed Ahmed_1
Occasional Visitor

Re: 10gR2 RAC on HP UX IA 11.31

Hi Duncan,
We have added additional memory of 3GB on first node and 6GB on second node on a temporary basis and not sure if we will get the exact results for vmstat and swapinfo
Also to add even CPU util goes close 100%

On first node:
$ vmstat 2 10
procs memory page faults
cpu
r b w avm free re at pi po fr de sr in sy c
s us sy id
11 1 0 666450 147618 0 0 6 0 0 0 5 3730 99464 494
7 37 19 44
11 1 0 666450 147523 0 0 0 0 0 0 0 2790 72879 393
4 17 5 77
7 1 0 702930 147521 0 0 0 0 0 0 0 2915 65477 398
8 29 8 63
7 1 0 702930 147507 0 0 0 0 0 0 0 2562 56282 338
1 42 18 40
7 1 0 702930 149385 0 0 0 0 0 0 0 2365 52133 304
7 49 1 50
7 1 0 702930 149365 0 0 0 0 0 0 0 2152 43206 267
4 11 9 80
7 1 0 702930 149365 0 0 0 0 0 0 0 1722 34156 217
4 7 3 90
14 1 0 739536 144859 0 0 1 0 0 0 0 1424 32687 181
6 33 39 28
14 1 0 739536 144915 0 0 0 0 0 0 0 2822 48863 434
4 33 15 52
14 1 0 739536 144825 9 0 0 0 0 0 0 2693 45535 388
0 27 27 46

$ /usr/sbin/swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 6144 168 5976 3% 0 - 1 /dev/vg00/lvol2
reserve - 5967 -5967
memory 8612 6001 2611 70%
total 14756 12136 2620 82% - 0 -

$ /usr/sbin/kctune | grep filecache
filecache_max 903068876 10% Imm (auto disabled)
filecache_min 451534438 5% Imm (auto disabled)

For Node2:
$ vmstat 2 10
procs memory page faults cpu
r b w avm free re at pi po fr de sr in sy cs us sy id
15 1 0 570043 979380 0 0 4 2 0 0 14 3696 88652 4522 31 19 50
11 1 0 513410 981098 0 0 2 0 0 0 0 5552 85159 7800 55 31 14
11 1 0 513410 981718 0 0 0 0 0 0 0 4944 153446 6850 62 37 0
11 1 0 513410 979853 0 0 0 0 0 0 0 3560 203221 5201 42 58 0
11 1 0 513410 979654 0 0 0 0 0 0 0 3824 175436 5509 54 28 18
11 1 0 513410 984159 0 0 0 0 0 0 0 4218 136360 5887 46 20 34
10 1 0 577960 982531 0 0 0 0 0 0 0 4137 124085 5763 49 31 20
10 1 0 577960 979658 0 0 0 0 0 0 0 3376 94295 4766 43 45 11
10 1 0 577960 987332 0 0 0 0 0 0 0 4526 90367 6560 46 19 34
10 1 0 577960 985892 0 0 0 0 0 0 0 4557 80252 6259 19 14 67

$ /usr/sbin/swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 6144 248 5896 4% 0 - 1 /dev/vg00/lvol2
reserve - 5893 -5893
memory 11640 5746 5894 49%
total 17784 11887 5897 67% - 0 -
$ /usr/sbin/kctune | grep filecache
filecache_max 1220566220 10% Imm (auto disabled)
filecache_min 610283110 5% Imm (auto disabled)


Thanks
Naveed Ahmed