From: "Carlo Stonebanks" <stonec.register@sympatico.ca>
Subject: Re: Performace Optimization for Dummies
Date: Tue, 3 Oct 2006 05:32:24 -0400
Organization: Hub.Org Networking Services
Lines: 23
Message-ID: <eftand$g5k$1@news.hub.org>
References: <efgu4v$276m$1@news.hub.org> <1159527507.2767.294.camel@holly>
	<efsjpk$1cb7$1@news.hub.org> <45221D5D.2020003@logix-tt.com>
To: pgsql-performance@postgresql.org

> I still think that using a PL in the backend might be more performant
> than having an external client, alone being the SPI interface more
> efficient compared to the network serialization for external applications.

I would actually love for this to work better, as this is technology that I 
would like to develop in general - I see db servers with strong server-side 
programming languages as being able to operate as application servers, with 
the enterprises business logic centralised on the server.

The import routine that I wrote will actually work on the server as well - 
it will detect the presence of the spi_ calls, and replace the pg_* calls 
with spi_* calls. So, you see this WAS my intention.

However, the last time I tried to run something that complex from the db 
server, it ran quite slowly compared to from a client. This may have had 
something to do with the client that I used to call the stored procedure - I 
thought that perhaps the client created an implicit transaction around my 
SQL statement to allow a rollback, and all of the updates and inserts got 
backed up in a massive transaction queue that took forever to commit.

Carlo