  | | | swap size recommendation for large DB server | swap size recommendation for large DB server 2006-08-11 - By Clark, Patricia
Back -- --Original Message-- -- From: taroon-list-bounces@(protected) [mailto:taroon-list-bounces@(protected)] On Behalf Of Stanley, Jon Sent: Friday, August 11, 2006 7:51 AM To: taroon-list@(protected) Subject: swap size recommendation for large DB server
I have a relatively large database server (quad Opteron with 32GB RAM). Oracle is having some type of performance issue on this host.
Oracle support is saying that we need swap that's 2x physical RAM. With todays systems that have huge amounts of memory, that's just not feasible, practical, or desirable from a SA's standpoint.
What I'm looking for is some BCP about how to size swap with large amounts of RAM. I know that it doesn't make sense, but I need something to throw in Oracle's face to say their recommendation is bogus.
BTW, the system isn't even touching swap.
-- Typical answer from a first line support person and it's incorrect. Recommendations are 2X RAM upto 2 GB of real memory + 1X RAM after that. It is also recommended that there are multiple SWAP files (partitions) when they get fairly large...I've seen 2 GB size recommendations.
That being said, swap is not the issue with oracle. The amount of shared memory that is set up in /etc/sysctl.conf in addition to the parameters that your DBA(s) use in their instances are more important. The oracle installation doco for linux has the recommendations for /etc/sysctl.conf . The rest is up to your DBAs. Unfortunately, our linux installs are rather small DBs so I don't have more specifics. I do know that our big production DB on Sun ran into these types of issues because Sun defaults to 8 GB of process memory allocation which required OS parameter tweaks, followed by DBA parameter tweaks.
Patti Clark Unix System Administrator - RHCT Office of Scientific and Technical Information
-- Taroon-list mailing list Taroon-list@(protected) https://www.redhat.com/mailman/listinfo/taroon-list
|
|
 |