RE: Filesystem Block Size ? 1K

From: Michael McMullen <ganstadba_at_hotmail.com>
Date: Thu, 16 Oct 2008 07:46:13 -0400
Message-ID: <BAY141-DAV118B52160622B710797E5BA6330@phx.gbl>
Message-ID: <F626BA5035746548AF2DF3E71FEB5257220877BB17@MBX01.bell.corp.bce.ca>


I've hesitated going to ASM for the reasons Jared listed and also because my SA's had such a hard time getting their heads around it. I'm using it on a new server that we have hooked up to a Pillar Data storage system. www.pillardata.com . Pillar lets you have a 1MB stripe size and as this matches ASM I went with this (10.2.0.4). In my very preliminary testing I'm getting the same performance in writes to the database as a straight up cp from one file to another. Does this make sense? Ie. I can copy a file at 183MB/sec and I can also get sqlloader at the same rate. I've never had this match up in any of my other systems. It was only after digging around that I figured out Pillar had the 1MB solution. Before that we were just using a regular filesystem. My speed there was 90MB/sec. To be fair, I haven't fully tested the filesytemio_options as I think I need it at direct to compare apples to apples.
There is a learning curve with ASM, for instance, my first db creation failed so I had to go into ASM and delete the controlfiles. That was a hassle and I have a lot more learning to do but it has been enjoyable to play with something new.
Pillar also has a cool EM plugin that lets you see everything which makes it a lot easier to view performance specs.

I have another question with ASM but I think I'll post it as a new thread.

--
http://www.freelists.org/webpage/oracle-l
Received on Thu Oct 16 2008 - 06:46:13 CDT

Original text of this message