Re: PostgreSQL vs Oracle

On Sun, Dec 21, 2008 at 1:17 AM, Victor Nawothnig
<victor(dot)nawothnig(at)gmail(dot)com> wrote:
> Hi,
>
> I am looking for some recent and hopefully genuine comparisons between
> Oracle and PostgreSQL regarding their performance in large scale
> applications. Tests from real world applications would be preferable
> but not required. Also differentiations in different areas (i.e.
> different data types, query structures, clusters, hardware, etc.)
> might be helpful as well.

Due to the terms of the license for Oracle, no one can publish
benchmarks without their permission.

Having used both Oracle 9 and pgsql from 6.5 to 8.3, I can say that
PostgreSQL is competitive for most small to medium loads I've thrown
at it, and given the high cost of licensing for oracle, you can throw
a LOT of hardware at PostgreSQL to catch up the last 10 or 20%
slowdown you might see in some apps.

Most of Oracle's advantages are in the more advanced features like
partitioning and reporting functions.

That said, I find PostgreSQL takes a LOT less administration to keep
it happy. Oracle doesn't just allow a wider range of performance
tuning, it demands it. If you don't demonstrably need Oracle's
advanced features then PostgreSQL is usually a better choice.

Last place we worked we developed on pgsql and migrated to oracle in
production (this was in the 7.4 era, when Oracle 9 was noticeably
faster and better than pgsql for transactional loads.) It was very
easy to write for pgsql and migrate to oracle as most SQL queries
didn't need any serious changes from one db to the other.

PostgreSQL generally tries to follow the SQL specs a little closer,
oracle has more crufty legacy stuff in it.

So, when you say large scale applications, are you talking OLTP or
OLAP type workloads? My experience has been that very few OLTP apps
get very large, as they get partitioned before they get over a few
dozen gigabytes. OLAP, OTOH, often run into hundreds of Gigs or
terabytes. I've found pgsql competitive in both really.