views:

149

answers:

3

I'm wondering about how to scale a database. Currently it uses PostgreSQL. Would switching to Oracle be worthwhile inspite of the coding pain and expense? Or is PostgreSQL + more boxes a better/cheaper approach?

+6  A: 

IME there's not a lot between any of the major DBMS in terms of the performance that you can achieve by performance tuning. It's not just my experience Open Source PostgreSQL Trails Oracle In Benchmark, But Not By Much

There is no substition for finding any performance problems and fixing them both in terms of client side and database optimisiation.

Richard Harrison
+1 for focusing on the existing problems rather than creating a whole bunch of interesting new ones.
APC
After working full time with Oracle and Postgres and running similar loads, if there's a difference - it just doesn't feel like much difference. +1 for not opening up a whole new mess.
rfusca
Note that the benchmark results Richard linked to here were using PostgreSQL 8.2; version 8.3 had a very large performance gain compared to earlier versions, in pretty much every way. Oracle may have improved some, too, but PostgreSQL is commonly 30 to 50% faster now than version 8.2 was.
Greg Smith
+2  A: 

I think that what you're suggesting is logically equivalent to Things You Should Never Do, Part I. This type of conversion is effectively a rewrite, and converting from one database to another (or one language to another, etc) is a great way to kill a lot of time and money that could otherwise be spent on actually improving your product. A better choice, as @Richard Harrison said above, would be to spend some time fixing whatever performance issues are driving you to consider moving to Oracle. As Fred Brooks said many years ago, there is No Silver Bullet.

Share and enjoy.

Bob Jarvis
A: 

I'll admit up front that I'm biased here, but I gotta tell you PostgreSQL is an amazing product! I'll admit I don't have any direct experience with Oracle, but everything I've ever read puts PostgreSQL within such a close range that even if it is a bit slower there is no question on my mind that spending any amount you could spend on licensing for Oracle instead on hardware or additional development time (to improve performance) should still leave you ahead in both dollars and performance.

Of course I do stipulate as others above have said that this is a question that really is too subjective unless we are going to talk about a particular situation or installed environment.

a2j