concurrent i/o

From: Jeff Barratt (jbarratt@COMPSAT.COM)
Date: Wed Oct 08 2003 - 08:48:12 EDT

• Next message: Nita Hyde: "Re: How do I drop raw devices/lv?"
• Previous message: Bill Verzal: "Re: How do I drop raw devices/lv?"
• In reply to: Clifton, Pablo: "Re: P690 and Shark Performance Problem"
• Next in thread: Jeff Barratt: "Re: P690 and Shark Performance Problem"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

I am contemplating using concurrent i/o with oracle database files

a'la
mount -o cio /u0#

specifically under AIX 5.2ML1 64bit kernel JFS2(not inline logs)running
oracle 8.17.3-32bit (will be moving to 9.1 64bit)

does any have any hard and fast gotchas with concurrent i/o ? I know that
concurrent i/o has all of the failings of direct i/o but supposedly the
inode locking is handled much more efficiently and performance studies(
http://www-1.ibm.com/servers/aix/whitepapers/db_perf_aix.pdf) show it to be
nearly equivalent(<10% slower) to raw device speeds.

• Next message: Nita Hyde: "Re: How do I drop raw devices/lv?"
• Previous message: Bill Verzal: "Re: How do I drop raw devices/lv?"
• In reply to: Clifton, Pablo: "Re: P690 and Shark Performance Problem"
• Next in thread: Jeff Barratt: "Re: P690 and Shark Performance Problem"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 22:17:16 EDT