Performance Issues on HP-UX 11.31, ia64 hp server rx6600

[Abhzi]

Hello All,

We are facing lot of performance issues on our database server. I have been told to check if system can be more tuned so that load can be normal. I am looking for the changes that I can make into the system so that more resources can be avialable for ORACLE processes to run or if some parameters that I can tune.

Below are details of the server, Please do let me know if any logs or ouputs are required. Thanks

=========

bash-4.2# dmesg | grep -i physical
    physical page size = 4096 bytes, logical page size = 4096 bytes
    Physical: 83854508 Kbytes, lockable: 62907572 Kbytes, available: 71533840 Kbytes

bash-4.2# swapinfo -tam
             Mb      Mb      Mb   PCT  START/      Mb
TYPE      AVAIL    USED    FREE  USED   LIMIT RESERVE  PRI  NAME
dev       32768     399   32369    1%       0       -    1  /dev/vg00/lvol2
dev      135168       0  135168    0%       0       -    3  /dev/vg03/lvswap
reserve       -   18638  -18638
memory    77889   52542   25347   67%

bash-4.2# machinfo
CPU info:
  4 Intel(R) Itanium 2 9100 series processors (1.59 GHz, 18 MB)
          532 MT/s bus, CPU version A1
          8 logical processors (2 per socket)

Memory: 81889 MB (79.97 GB)

Firmware info:
   Firmware revision:  04.15
   FP SWA driver revision: 1.18
   IPMI is supported on this system.
   BMC firmware revision: 5.26

Platform info:
   Model:                  "ia64 hp server rx6600"
   Machine ID number:      c2014b17-0bbc-11de-96b3-70aa34696848
   Machine serial number:  DEH49056X0

OS info:
   Nodename:  oracledb2
   Release:   HP-UX B.11.31
   Version:   U (unlimited-user license)
   Machine:   ia64
   ID Number: 3254864663
   vmunix _release_version:
@(#) $Revision: vmunix:    B.11.31_LR FLAVOR=perf
bash-4.2#

[Dennis Handly]

>dev      135168       0  135168    0%       0       -    3  /dev/vg03/lvswap

You have way too much device swap.  If you need that much, you'll have performance problems.

Do you have glance installed?  Does top point to any problems?

[Abhzi]

Thanks Dennis for your reply and apology for the late response.

No, glance is not installed.

At times server loads goes very high during production hours and the server becomes very slow. I can see from the top output at that time, all major processes are only of oracle. But my query is to do some parameter tuning?  i.e. to tune some parameters at OS level so that more resources will be available for database to use or to do some kind of load balancing from the processors side( to ideally share the load). So that there more processes can be accommodated keeping load not much high.

[Bill Hassell]

All performance issues require a lot of work to resolve -- there is no GO-FASTER parameter in HP-UX. In fact, if you stop running Oracle, your system will probably run very fast. So before looking for a secret parameter, your database admin needs to run an Oracle statspack analysis to determine if there more SGA memory is needed, more indexes, rewriting inefficient SQL queries, etc. Without Glance, it is very difficult to determine where the limitations are occurring. It could be that your system is running at 100% CPU usage (a good thing but if this seems slow, then you need to double or triple the number of CPUs in your system.And if you are also running JavaJunk on the system, have your Java admin look at optimizing the code -- Java can be a huge memory and CPU hog.

Or you could be hitting a limit for disk I/O. And there are two causes: bad Oracle design that is using more disk I/O than necessary, or your disks are simply too slow for this application. If your disks are JBODs (no smart array controller) then adding additional disks and paths and spreading the data across more spindles might help. If you are using a smart array controller, make sure you have all the cache possible and that your storage administrator has analyzed I/O and LUN assignments for performance.

[chris huys_4]

Hi,

Bills oracle statspack, oracle awr is it called now, is a good idea. Do it during a "good" and a "bad" period, during a 10-15 minutes period. Unless you only have bad periods.

For the rest exuecute on seperate terminals  sar -u 1 2000 and sar -LdR 1 2000 output taken during the above "good" and "bad" period oracle awr taking.

Then log a support call with hp and ask them , to ask you, for kitrace output.

And if you want to tune something upfront.

See if the advises in the technical whitepaper for vxfs mount options for oracle databases were implemented.

HP-UX VxFS mount options for Oracle Database environments

http://h20195.www2.hp.com/V2/GetPDF.aspx/4AA1-9839ENW.pdf

(i.e. implementing direct IO through delaylog,mincache=direct,convosync=direct, which will bypass the hp-ux os buffercache/filecache,  for filesystems with oracle archive logs, oracle redologs and oracle tablespaces files)

Best Regards,

Chris