Tom Duerbusch wrote:
My initial assumption about this is:
On Intel/AMD, they are using 512 byte blocks. Standard PC file system...right?
The _disk_ might use 512 byte sectors, but the filesystem does not. Odds
are good your very expensive DASD also uses 512 byte sectors.
How the data gets from user-space to disk (and even whether it does) is
up to the kernel. I'd not expect the application tools to be able to
measure the real disk I/O as it applies to a particular application.
And even though we are using the same 512 byte blocks under the RAID covers, we
are using, at least 4K blocks if not tracks.
Mark didn't say how Oracle's managing storage on the toy. I'd want to
know it's using properly-equivalent specifications: if it's using files
on one and raw disk on the other, all bets are off.
I'd not consider reiserfs on one, ext3 on the other to be equivalent either.
So, on each of the Linux images, you need to look at:
/proc/dasd/statistics
To see what you are dealing with.
See
http://www.ibm.com/developerworks/linux/linux390/perf/tuning_how_tools_dasd.html
to help understand /proc/dasd/statistics.
Tom Duerbusch
THD Consulting
Mark Post <[EMAIL PROTECTED]> 6/13/2007 3:26 PM >>>
I got an internal query about one of our customers who is currently running
Oracle on Linux on Intel/AMD. They're looking at moving that to Linux on the
mainframe, and they've seen some good things during their testing. For
example, one long-running query went from 14+ minutes to about 2.5 minutes.
Something that caught their attention though was that the number of I/Os on the
mainframe were about 30 times (not percent) higher than on Intel. They said
they'd used an Oracle tool to determine this. Not knowing anything about
Oracle's tools in this arena, I have no idea if they're any good or not. I
told them they needed to get a real performance monitor to verify that, but
with a 5.6-to-1 performance _improvement_, it's not yet critical.
My question is, does anyone know if Oracle uses different internal algorithms
on the various architectures to maximize performance? That is, since I/O is a
strength of the mainframe, they might do more I/Os versus something else on
Intel Linux. Anyone have any insight on this?
--
Cheers
John
-- spambait
[EMAIL PROTECTED] [EMAIL PROTECTED]
Please do not reply off-list
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
