Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Performance difference between 2 machines

Re: Performance difference between 2 machines

From: <xhoster_at_gmail.com>
Date: 21 Jun 2005 00:22:11 GMT
Message-ID: <20050620202211.387$Pm@newsreader.com>


"Joel Garry" <joel-garry_at_home.com> wrote:
> Something is bothering me about all this. Here is the user guide:
> http://h200005.www2.hp.com/bc/docs/support/SupportManual/c00311965/c00311
> 965.pdf See the instructions for checking the array controller logs.
>
> With a RAID setup spreading the I/O among many spindles, why would
> rotational delay be such a big deal with 2 I/O's given sequentially at
> the same time through the OS level be different than 2 I/O's given
> through the Oracle level to the OS level?
>
> I have to wonder what else is going on. Would the OP have glance
> available? Is db_file_multiblock_read_count set already? Is the
> pitiful server using a much faster cpu than the hp? Did the test get
> done multiple times? Is it configured for Autoraid (those would be
> 256K blocks...)? Stripe size? Watch for swapping?
>
> The customer machine had fewer, much longer waits than the pitiful
> machine. Almost as though they lost their place in line for some
> reason.
>
> There's something I'm missing here.

The 0.5 ms average wait time for the faster system made me wonder if it was reading from OS/FS cache instead of from disk.

Xho

-- 
-------------------- http://NewsReader.Com/ --------------------
Usenet Newsgroup Service                        $9.95/Month 30GB
Received on Mon Jun 20 2005 - 19:22:11 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US