RE: RE: 11g RedHat 5 and Hugepages

Possibly- there are issues with MTS and they want enough headroom so
that they can run without MTS and possibly connection pooling tuned very
conservatively on the app host side.  On the current, undersized
configuration, they have already seen 75,000 connections at peak
periods.

 

From: Niall Litchfield [mailto:niall.litchfield@xxxxxxxxx] 
Sent: Wednesday, October 13, 2010 11:24 AM
To: Crisler, Jon
Cc: Oracle-L@xxxxxxxxxxxxx; ora_kclosson@xxxxxxxxx
Subject: Re: RE: 11g RedHat 5 and Hugepages

 

Sheesh. You normally can get between 10 and 50 connections per pooled
connection - so you seem to be looking in the order of several million
concurrent connections across the cluster?

        On 13 Oct 2010 16:13, "Crisler, Jon" <Jon.Crisler@xxxxxxx>
wrote:

        In my case, I am looking at a connection count of at least 5000,
with an SGA of 70gb or higher, on each node of a 6 node cluster, with
each node having 48 cpu's.  We have run into issues with MTS and will be
using dedicated connections on the DB side but the numerous app servers
have connection pooling.  The database size will be about 6 TB to start.
Preliminary testing on a small system already showed a performance gain
when using hugepages, but the test was on SUSE Linux while we are using
RedHat 5, so its not a 100% accurate comparision.

         

        This 6 node cluster will replicate to an identical cluster with
DataGuard.

         

        From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Kevin Closson
        Sent: Wednesday, October 13, 2010 10:56 AM
        To: Oracle-L@xxxxxxxxxxxxx

        
        Subject: Re: 11g RedHat 5 and Hugepages

        
        
         
        
        Nothing wrong with this feedback because it includes information
about how to do it right (e.g....

Other related posts: