Re: XMLType performance

From: Serge Rielau <srielau_at_ca.ibm.com>
Date: Thu, 12 Apr 2007 13:27:18 -0400
Message-ID: <5878fnF2eab6aU1@mid.individual.net>

Brian Peasland wrote:

>> When a schema is highly volatile, e.g. tax-return forms, XML beats 
>> relational hands down. Tables are so...square.
>> Also XML documents can contain electronic signatures. Shredding and 
>> reconstitution can cause serious headache to that end.
>> What I see is that XML is stored in the database, but the core 
>> reference  attributes are pulled out into regular column for quick 
>> processing (and RI enforcement, ..).
>>
>> The right tool for the right job.

> The XML structure is hierarchical in nature. Haven't we shown over the
> last umpteen years that hierarchical databases are have more than enough
> issues in which to make relational databases a much more attractive
> solution? There will always be the exceptions to the rule, but give me
> relational databases over hierarchical databases almost any day. To that
> end, why store XML data in XML form in a relational database? One is
> just forcing a hierarchical database into a relational model.
I think each of the major vendors will claim that they are not selling RDBMS anymore. The fact of the matter is that any major organization/comany today has a lot of relational data, a lot of unstructured data and a lot of semi-structured data. Oracle will be the first to argue that all the data belongs into Oracle and not Filenet or Tamino ;-)

BTW. IMS has healthy growth rates and runs circles around any RDBMS for many OLTP workloads. There is good reason why the world's banking still runs on that technology. It's good in what it does....

> Granted, if you are storing an XML document, then I can see the need to
> store this in an Oracle database no different than if I were storing a
> PDF or Word document in the database, as a single instance of an entity.
Yet there is a big push to peek into the content of the PDF document or word document I think "enterprise search" is the bang word here. Regular expression matching also caters to the not so relational aspects.

> And I only want to be querying for the entire document. I do not want to
> be pulling pieces out of that XML document. If I'm going to do that,
> then I'd much rather store the attributes in a relational table.
Let's go back to the tax forms: This attribute is a line in your tax-return. It was introduced in year x and it will disappear in year y depending on how the political wind blows. Now one can insist in relational by pivoting the tables and many products do just that. But what's relational about a table that adds "COLUMNNAME" to the primary key and an ANYTYPE column named "VALUE"?

Cheers
Serge

-- 
Serge Rielau
DB2 Solutions Development
IBM Toronto Lab

Received on Thu Apr 12 2007 - 12:27:18 CDT