Re: Implementing trees in a relational database
From: 9rowzn01i001 <9rowzn01i001_at_9rowzn01i001.com>
Date: Tue, 30 Jul 2002 09:32:18 +0000 (UTC)
Message-ID: <ai5mf1$g0n$1_at_venus.btinternet.com>
"Paul DeWolf" <paul_at_thievesandkings.com> wrote in message news:uLn19.179675$Wt3.134644_at_rwcrnsc53...
> Hi all,
>
> I used to work in the application development area of a very large company
> who built applications on top of very complex data models using a very
large
> relational database vendor.
>
> We had various networks (recursive many-to-many relationships) that were
> stored in the database through "connectivity" tables, but it was
*extremely*
> awkward to develop the software both on the application and database sides
> and it seemed that we had real scalability and performance problems,
> especially as we wanted to add to the data model.
>
> Since then I've moved to working for an object-oriented database company
> where these relationships are quite easy to develop and scale easily, but
> I'm still curious about how one can implement trees and networks in an
> RDBMS.
>
> Can someone point me to information (books, white papers) on techniques
for
> efficiently implementing trees and networks in an RDBMS? Has anyone done
it
> themselves and showed that they can scale linearly?
>
> Paul DeWolf
> Systems Engineer
> Objectivity, Inc.
>
>
>
Date: Tue, 30 Jul 2002 09:32:18 +0000 (UTC)
Message-ID: <ai5mf1$g0n$1_at_venus.btinternet.com>
"Paul DeWolf" <paul_at_thievesandkings.com> wrote in message news:uLn19.179675$Wt3.134644_at_rwcrnsc53...
> Hi all,
>
> I used to work in the application development area of a very large company
> who built applications on top of very complex data models using a very
large
> relational database vendor.
>
> We had various networks (recursive many-to-many relationships) that were
> stored in the database through "connectivity" tables, but it was
*extremely*
> awkward to develop the software both on the application and database sides
> and it seemed that we had real scalability and performance problems,
> especially as we wanted to add to the data model.
>
> Since then I've moved to working for an object-oriented database company
> where these relationships are quite easy to develop and scale easily, but
> I'm still curious about how one can implement trees and networks in an
> RDBMS.
>
> Can someone point me to information (books, white papers) on techniques
for
> efficiently implementing trees and networks in an RDBMS? Has anyone done
it
> themselves and showed that they can scale linearly?
>
> Paul DeWolf
> Systems Engineer
> Objectivity, Inc.
>
>
>
Maybe this will help, maybe it won't. There is a white paper by Jacques Roy (Informix, now IBM) that describes the Node Type that can be used to model hierarchical data:
http://members.lycos.co.uk/xmldatabases/misc/
There are some performance numbers too.
I know that Node has been implemented on Informix IDS, Cloudscape and DB2.
Paul Brown (contributes to comp.databases.theory) developed the Node Type and has also written a very nice chapter on how it can be used to manage XML in Object-Relational DBs for a forthcoming book I'm putting together.
vr
-- akmal chaudhri IBM developerWorks - http://www.ibm.com/developerWorks/ XMLDatabases ------- http://www.btinternet.com/~xmldatabases/Received on Tue Jul 30 2002 - 11:32:18 CEST