Re: Orbitz dumps RAC for better reliability

From: Mark D Powell <Mark.Powell_at_eds.com>
Date: 18 Jul 2003 05:50:39 -0700
Message-ID: <2687bb95.0307180450.c71eba6_at_posting.google.com>


Keith <nospam_at_nospam.com> wrote in message news:<vhf46a464qn62_at_news.supernews.com>...
> Stonebraker's original contribution is far different from today's
> PostgreSQL. Also it is based on postgres not Ingres. Ingres goes back to
> earlier work. Please research before you post.
>
> Note: good designs comes often come from academia or research
> institutions like Bell Labs. Relational Data Management came from an
> academic and researcher, E.F. Codd. Larry and his cohorts simple ripped
> off the research paper.
>
> In my opinion, Oracle today is simply an organization of Marketeers and
> salespeople.
>
> Nuno Souto wrote:
>
> > Orville <hawaiian_at_bacon.com> wrote in message news:<Xns93BBCFD8D2F68hawaiianbaconcom_at_216.148.227.77>...
> >
> >
> >>Truth in advertising.
> >
> >
> > Contradiction in terms?
> >
> >
> >
> >>Customer testimonials:
> >> "100% of the data was gone!"
> >
> >
> > What else? And look, it's not even clustered!
> >
> >
> >
> >>100% of the data was gone! From the shell if you went
> >>select * from company, postgres reports 0 tuples! (but returns
> >
> >
> > Look, this crap is based on Stonebraker's Ingres.
> > Like the original, these idiots concentrated on
> > "functionality" instead of making sure they had a
> > solid data management engine.
> > Like the original, it loses data at the drop of a hat.
> > Like the original, if a layer of serious data management
> > is added the performance goes out the window.
> >
> >
> >>Please not I was running fsync off for this database.
> >
> >
> > He can run fsync until the cows come home. The problem
> > is and has always been the total wanton disregard for
> > data security from the Stonebraker mob. They don't
> > have a SINGLE person with data management experience
> > ANYWHERE in their group! All a bunch of academics
> > with pretentious titles and even bigger egos, that's all.
> >
> >
> > Cheers
> > Nuno Souto
> > wizofoz2k_at_yahoo.com.au.nospam

Keith, E.F. Codd worked for IBM when he invented relational theory and relational algebra to define and explain his theory. Bell Labs in its hayday was ran as a business operation. Academia sometimes has some good basic ideas, but it almost always fails in any attempt to bring them out into the real world.

Also was not Postgres based on Ingres so if PostgresSQL is based on Postgres and Postgres was based on Ingres then it seems reasonable to say that PostgresSQL is based on Ingres. However, just because a piece of software is based on an earlier version does not mean it inherits all of the flaws or good features of the prior version. Code changes, sometimes for the better, and sometimes for the worse.

Getting back to Orbiz. I have yet to follow a link to an article that actually explains the problem actually encountered. One of the articles listed the May transaction numbers. If the system worked then, why is it failing now? What, if anything, changed? Without facts, speculation about the problem is just that, speculation.

As far as PostgreSQL being a good database. It may suitable for a lot of uses; however, the last I knew when you delete a row in PostgreSQL the row is actually marked as logically deleted and you have to run a utility which takes exclusive control of the target table to [Quoted] physically delete the rows and reclaim space. When the economy is good my client requires 24x6.9 access. The small window we get to move changes into production is way too short to clean up deleted data. Our applications require an rdbms that can handle concurrent insert, update, and delete access to the same table from multiple batch and online processes. Oracle provides that kind of access with good performance. When PostgreSQL can do that, I will consider it as an option.

  • Mark D Powell --
Received on Fri Jul 18 2003 - 14:50:39 CEST

Original text of this message