Re: [SQL] Re: [HACKERS] Proposed Changes to PostgreSQL - Mailing list pgsql-hackers

From Marten Feldtmann
Subject Re: [SQL] Re: [HACKERS] Proposed Changes to PostgreSQL
Date
Msg-id 200002041815.TAA07061@feki.toppoint.de
Whole thread Raw
In response to Re: [HACKERS] Proposed Changes to PostgreSQL  (Chris Bitmead <chrisb@nimrod.itg.telstra.com.au>)
List pgsql-hackers
> Peter Eisentraut wrote:
>
>
> This is a really stu^H^H^H bad idea. I have hierarchies 5 levels deep
> with
> multiple inheritance, and I
> don't want to do a 10 way join just to retrieve an object.
>
> This is why RDBMS's performance sucks so incredibly badly on some
> applications.
> an ODBMS can perform 100x as fast in these cases just because of what
> you
> are proposing.
>

 Hmm, and yes one may find problems where the pure relational system
is 100x faster than your ODBMS.

 After doing a project with VERSANT and VisualWorks (election projection
system for the first television sender here in Germany) I like the
idea of OODBMS, but I've also noticed, that they are not the solution
to all problems.

 Clever database desing leeds to good performance on both systems, but
one should consider, that the designs of the database layout will be
different. There are cases, where a pure relational system is very
fast and an ODBMS never get it, but there are the examples you
mentioned.

 Joins per se are not that bad .. it depends on when and how they
are used and how good the analyzer of the database is and how good
he uses the indices to get the job done.

 One very good point is the query language of the rdbms systems. On
the odbms side no standard is really available, which can be seen as
the sql of the odbms.

 Marten



pgsql-hackers by date:

Previous
From: Patrick Welche
Date:
Subject: Re: [HACKERS] Another nasty cache problem
Next
From: Michael Meskes
Date:
Subject: Re: [HACKERS] Two backends at the same time