Hey you shining happy *nin users,
Last night I attend an Oracle presentation at the UUASC group
(http://uuasc.org/oc.html) for an Oracle clustering presentation
(http://uuasc.org/p0611.html).
I've only done Windows Server 200x clustering for Exchange and SQL 2000,
so it was interesting to hear some of the considerations one has to take
into account for Oracle clusters and how various Sun utilities make
these deployments easier than, say, Red Hat.
For what it's worth, here were my questions for the speaker:
1. How does your database structure affect the way you carve up your
LUNs (e.g. an EMC)?
(I was hoping he'd expound on some of the RAID stuff I recently learned
from this MS book: http://tinyurl.com/yn6tyc)
He didn't go in depth as I had hoped, but he layed out a sample SAN and
showed how he'd RAID 10 it in 4 drive chunks and said that 6 drives (for
whatever reason) was optimal.
2. What tools to you use to help determine chokepoints?
He answered Sar and said that it's on almost every UNIX variant (a tool
I've never played with).
http://www.computerhope.com/unix/usar.htm
He also mentioned some other tools, such as LoadRunner
http://www.mercury.com/us/products/performance-center/loadrunner/
Fire it up at the end of month, record read DB transactions during the
day, then play it back with your various perf mon tools to find the
chokepoint.
Roger