Re: Prespawning processes on dedicated server (8i)

From: Holger Baer <holger.baer_at_science-computing.de>
Date: Wed, 29 Jun 2005 13:11:46 +0200
Message-ID: <d9tvlk$ouo$1@news.BelWue.DE>

Billy wrote:
> Holger Baer wrote:
>
>

>>If the application server uses connection pooling, then there shouldn't
>>be constantly new connections be made and old ones closed. Which in turn
>>should result in a small pool of connections.

>
>
> In theory, yes. In practise, not. Or at least not with the JDBC
> connections from app servers that I'm dealing with on a range of Oracle
> platforms.

My experience (and my understanding what a pool is) obviously differs from yours, so I think we both have a valid point there.

The applications I've come across build up a connection pool upfront, using dedicated server connections. The enduser sessions don't trigger new connections, and they certainly don't get closed after usage.

>
>

>>Pooling a pool is not a good idea in my book.

>
>
> Oracle itself does not care and does not know that this is a pooled
> connection that is established. There is *no* difference between such a
> connection and a "normal" connection from an Oracle perspective. So I
> fail to understand your concerns Holger.

Caching a cache. Not a good idea. Ok, there are rare occasions such as 2nd and 3rd level caches in CPUs where it can make sense (but often fails to really help), but with IO-Systems you never do that if you can avoid it. The same applies to MTS and Appserver connection pools (which in a way are IO-Systems).

Regards,
Holger Received on Wed Jun 29 2005 - 06:11:46 CDT