You're correct on trying and figuring the appropriate value of shared_buffers for our application which I'd be doing. However, I wanted to test how far I can get specially that 8.1 removed the limit of 2GB on a 64-bit architecture like ours. The other thought is that my assumption (which I may be wrong about) that the RDBMS (postgres in this case) will do a better job managing and prioritizing the caching of objects, ie. tables, data, etc than the kernel's cache which doesn't distinguish indexes from tables,etc. I might be wrong on that (Linux kernel), but that was my experience with Oracle on a Solaris box. Thanks, --- Husam Tomeh -----Original Message----- From: Tom Lane [mailto:tgl@xxxxxxxxxxxxx] Sent: Wednesday, December 14, 2005 2:38 PM To: Tomeh, Husam Cc: pgsql-admin@xxxxxxxxxxxxxx Subject: Re: [ADMIN] PG 8.1.1 Cannot allocate shared_buffers memory error "Tomeh, Husam" <htomeh@xxxxxxxxxxx> writes: > I just installed postgres 8.1.1 on a dual optron Linux box running Linux > 4 advanced server, with 32GB of RAM and 2GB of swap. > I'm trying to allocate 24 GB as shared_bufferes. I'm getting an error > that I can not allocate memory (see below). While this doesn't answer your question: why are you trying to do that? AFAIK there is no evidence whatever that that's a good idea. Try measuring performance at, say, 1Gb, 8Gb, 16Gb of shared buffers and see where the trend is going before you waste a lot of time fixing the kernel parameters. regards, tom lane ********************************************************************** This message contains confidential information intended only for the use of the addressee(s) named above and may contain information that is legally privileged. If you are not the addressee, or the person responsible for delivering it to the addressee, you are hereby notified that reading, disseminating, distributing or copying this message is strictly prohibited. If you have received this message by mistake, please immediately notify us by replying to the message and delete the original message immediately thereafter. Thank you. FADLD Tag **********************************************************************