From: Chris Browne <cbbrowne@acm.org>
Subject: Re: [pgsql-advocacy] [PERFORM] Postgres VS Oracle
Date: Tue, 19 Jun 2007 09:39:02 -0400
Organization: cbbrowne Computing Inc
Lines: 48
Message-ID: <60ps3sdpxl.fsf@dba2.int.libertyrms.com>
References: <c50dbb0b0706180855v4b781f13m6e54c50aa3ab1dee@mail.gmail.com>
	<36e682920706180910y519bc6a5h11e0ba56f45009ec@mail.gmail.com>
	<4676B7C7.6020801@kostyrka.org>
	<36e682920706181002i495f3fc3s2d55472e36a91b63@mail.gmail.com>
	<4676BE31.6040701@commandprompt.com>
	<115F286D-0396-44DB-8E3B-8801B0924D66@indiana.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.20 (linux)
Cancel-Lock: sha1:upozoBoj4IIsmUkgilrX2FHLSNA=
To: pgsql-admin@postgresql.org

walterc@indiana.edu (Carol Walter) writes:
> I don't want to add gas to the flamewar, but I gotta ask.  What is in
> the the 90 to 95% referred to in this email.

I'd say, look at the Oracle feature set for things that it has that
PostgreSQL doesn't.

Four that come to mind:

- ORAC = multimaster replication
- Integration with hardware vendors' High Availability systems
- Full fledged table partitioning
- Windowing functions (SQL:2003 stuff, used in OLAP)

These are features Truly Needed for a relatively small percentage of
systems.  They're typically NOT needed for:

 - departmental applications that operate during office hours
 - light weight web apps that aren't challenging the limits of
   the most expensive hardware
 - any application where reliability requirements do not warrant
   spending $1M to make it more reliable
 - applications that make relatively unsophisticated use of data
   (e.g. - it's not worth the analysis to figure out a partitioning
   design, and nobody's running queries so sophisticated that they
   need windowing analytics)

I expect both of those lists are incomplete, but those are big enough
lists to, I think, justify the claim, at least in loose terms.

The most important point is that third one, I think: 
  "any application where reliability requirements do not warrant
  spending $1M to make it more reliable"

Adopting ORAC and/or other HA technologies makes it necessary to spend
a Big Pile Of Money, on hardware and the humans to administer it.

Any system whose importance is not sufficient to warrant *actually
spending* an extra $1M on improving its reliability is *certain* NOT
to benefit from either ORAC or HA, because you can't get any relevant
benefits without spending pretty big money.  Maybe the number is lower
than $1M, but I think that's the right order of magnitude.
-- 
output = reverse("ofni.secnanifxunil" "@" "enworbbc")
http://linuxdatabases.info/info/nonrdbms.html
"One disk to rule them all,  One disk to find  them. One disk to bring
them all and in the darkness grind them. In  the Land of Redmond where
the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh