Operating System - HP-UX
1825766 Members
2199 Online
109687 Solutions
New Discussion

how to check which process increase my I/O value

 
ng_7
Regular Advisor

how to check which process increase my I/O value

hi,

My top results shows that my cpu and memory still at acceptable level. but the result for sar -udd shows very high value as per attached. May I know which process caused this problem ?

thanks

ng
6 REPLIES 6
Hein van den Heuvel
Honored Contributor

Re: how to check which process increase my I/O value

Those IO stats do not look too bad.
They could easily be explained by a process scanning a large file, a backup operation, an import/export or such. The average queue suggests it could just be 1 process requesting the IO. The system is working. Be happy!

Other then GlancePlus there are, best I know, no standard tools to list high hitting IO user. --> proclist --> o --> 1 --> "Disk I/O Rate threshold".
That will require the problem (if there is a problem!?) to be actively present.

I think there are freeware tool to help (PIO, TOPIO) also, but often a simple 'top' will lead to a suspect process.

Be sure to read up:
- HPUX Tuning and Performance Manual (Sauers)
- HPUX Performance Cookbook (Ciullo)
- Tool list: http://docs.hp.com/en/5991-7469/5991-7469.pdf

I can not readily find a link to the cookbook.pdf on hp.com. Google find a link under dspp which does not work: http://h21007.www2.hp.com/dspp/files/unprotected/devresource/Docs/TechPapers/UXPerfCookBook.pdf

Html version: http://216.239.51.104/search?q=cache:I_YlSXKonuUJ:h21007.www2.hp.com/dspp/files/unprotected/devresource/Docs/TechPapers/UXPerfCookBook.pdf+hpux+Ciullo+cookbook+%2Bsite:hp.com&hl=en&ct=clnk&cd=1&gl=us

Peoplesoft copy: http://www.peoplesoft-hp.com/tools/whitepapers/%5Bfiles%5D/Technical/HPUX%20Performance%20Cookbook.pdf

Hope this helps some,
Hein van den Heuvel (at gmail dot com)
HvdH Performance Consulting



Novonil Choudhuri
Frequent Advisor

Re: how to check which process increase my I/O value

I donot think you have faced any performance issues due to large disk utilization. In fact sar -d output shows that your avwait values are pretty smaller compared to avserv.. so you need to worry..

Only when you see avwait values higher that avserv you need to check which process eating up your resources... :)
ng_7
Regular Advisor

Re: how to check which process increase my I/O value

but my Oracle system somehow is very slow until we need to reboot server . and after reboot server, the performance will be better but is only can last for 2 to 3 hours
Novonil Choudhuri
Frequent Advisor

Re: how to check which process increase my I/O value

ohhh..

so i advice you to log a call with HP support as well as Oracle.. Y

ou need to provide them glance outputs to HP for a detailed analysis....

Best of Luck!! :)
Mario_88
Advisor

Re: how to check which process increase my I/O value

The sar times don´t seems too bad. Apart from adding more disks and stripe the datas in them you can try to use direct IO.

If you have enough memory I would check the recommendation ov V$DB_CACHE_ADVICE

When does the performance increase, when you restart the oracle instance or when you reboot the system?

Hein van den Heuvel
Honored Contributor

Re: how to check which process increase my I/O value

>> but my Oracle system somehow is very slow until we need to reboot server . and after reboot server, the performance will be better but is only can last for 2 to 3 hours

You never told us this was Oracle.
Did you consult with your DBA peers?

When you indicate 'performance is better', how is that measured. End user response time? End user throughput? system load?
Sometimes folks confuse a low utilization with better performance... au contraire!

If you just restart Oracle, not the whole box, is the performance better at first also?

This could easily be a memory problem.
The Oracle memory SGA + PGA is not over committed compared to physical memory is it now? The SYSTEM cpu time component is higher than I would accept it on my system.

Oracle has great tools built in to explain what are top wait events, where its Os are going and so on: statspack, enterprise manager and so on.

For better help, you may want to summarize some statspack data in an attached TEXT file (no .doc / .rtf please!), and an indication of memory usage, and maybe label those drives as to their major function (Oracle data, redo, swap,...).

Finally, If this is a serious problem for the business, then you may want to consider some hands-on help. Be sure to check with peers around you, and/or request some professional consulting (like myself :-).

Hope this helps some,
Hein van den Heuvel (at gmail dot com)
HvdH Performance Consulting