Re: Real World Experience of Oracle RAC and its ability to Scale

Kent Stroker <kent_at_strokermedia.com> wrote in message news:<BE8B0836.AC6%kent_at_strokermedia.com>...

> If for no other reason than running the middle-tier and the database on the
> same hardware is just plain wrong; certainly will invalidate any perceived
> HA gains from RAC.

I've always wondered about this. Is it really that wrong to run the database and application server on the same hardware? I can see a few solid arguments against this -- OS level monitoring tools make it hard to differentiate the impacts, also the resource needs of an application server tend to differ from the DB, leading to harder to predict loads.

But as for HA, I can't see any general problems -- a node going down can be handled by the application server failover. A DB process-level failure can be handled by the database driver.

But what's the benefit of sharing an application server and database on the same machine? The only one I can really come up with is the reduced latency by using the IPC protocol in the TNS listener. For certain applications, with low CPU but high I/O requirements, that might be worthwhile. It all depends on your tests. And whether you can deal with the more complex monitoring & potential for contention.

Cheers
Stu Received on Wed Apr 20 2005 - 22:34:45 CDT