[ Cc: list cleaned a bit ]
On Thu, 20 Jan 2005, Alvaro Herrera wrote:
> On Thu, Jan 20, 2005 at 08:03:42AM -0500, Troyston Campano wrote:
>> Basically, we want to take 3 of the 10 applications running on Oracle, move
>> them to Postgresql on the same computer/server and just make sure it runs
>> about the same (really speed, memory usage, and space are the big issues).
>> I'm not concerned with how hard the migration will be and things like that.
>
> So you want Postgres to be a cheaper Oracle. Hmm. Maybe it will work,
> but as Marco Colombo says, you are not going to see Postgres shining by
> using that simplistic approach. If you want that, maybe you should look
> closely to see where you can find the rusty corner that needs to be
> polished.
>
>> The database is very low in complexity so the migration should be cake.
>
> If you do things as simple as "select count(*) from table", then you
> will have to be careful to be really fair in your comparison; you could
> misleadingly find that Postgres is much slower on that query.
Yeah, that's precisely what I meant. The Oracle to PostgreSQL migration
is well worth considering a (partial) redesign.
Sometimes I happen to show some SQL queries I make (on PostgreSQL)
to an Oracle guy. Usually it goes like this:
1) hmm (ponders at the query)
2) ah!!! (realizes what the query does)
3) I didn't know you could do this! (excitement)
4) hmm (ponders again if that may work on Oracle as well)
5) no I can't do that in Oracle that way. I remember I did something
like that in the past with other 3 SQL developers. We had to write
a 300 lines long stored procedure, we got the locking right at the
third reimplementation only (we don't know if we do need the locking,
we asked our senior DBA but he was unsure as well, so we put it in,
we don't think it hurts anyway).
Ok, point 5) has been exaggerated to joke level, but you get the idea.
.TM.
--
____/ ____/ /
/ / / Marco Colombo
___/ ___ / / Technical Manager
/ / / ESI s.r.l.
_____/ _____/ _/ Colombo@ESI.it
