Re: Poor Performance on Oracle9i - 9.2.0.7

eavanzi wrote:
> Hello, Beko.
> The results are:
>
> jobq slave wait 14425
> 3604979
> db file sequential read 20479048
> 4342690

This means you are doing a bunch of random gets. It is possible that this could be improved by the things I mentioned in a previous post (like multiple buffer pools). Does top show a lot of cpu spinning? Hot blocks in the SGA could mean the cpu is spending a lot of time just moving things about. Or it could just mean you have an I/O problem, saturation or contention. Did you say the redo and archiving are RAID-5?
> rdbms ipc message 313526
> 9989699

do an ipcs in the shell, be sure the things that should be there are there and in as few segments as possible. One thing that can happen is segments can be orphaned, slowing everything down by simply using up memory (I've seen cases where a reboot is required to get rid of them, on modern hp-ux and older slolaris. I don't recall if your OS requires removal of files that track these segments in the case of an instance crash.) Also, using a large SGA with 32 bit can cause a lot of windowing issues. Having been 64 bit for a long time, I don't recall much of that, but poke about on metalink.

> pipe get 38824
> 13604893

I have a feeling that this is bad, but don't remember why. I think it means tune your app, if your app is defining its own pipe then it may need to define it larger. I think it may also mean oracle processes are spending a lot of time communicating with each other, are you seeing lots of cpu with smon or pmon or arch? Do you have OEM performance manager available?

> SQL*Net message from client 10333655
> 434317043
>

You should figure out if this is just users thinking about what keys they want to hit or if there is a network problem. Are you using Forms?

jg

--
@home.com is bogus.
http://www.0lll.com/archgallery2/mvrdv_mirador/index.htm