Re: database performance questions

From: Jerry Gitomer <jgitomer_at_erols.com>
Date: 2000/02/04
Message-ID: <389A6DE4.6EC2E76_at_erols.com>#1/1


Kreiner Karl wrote:
>
> Hi everone,
> I'm currently planning a database architecture , that administrates
> documents . The structure ( properties ) of such documents may vary, so I
> decided to build an architecture that makes it possible for the user to add
> or change properties.
> The properties may be numeric, strings or date-time values, so I currently
> decided to build several tables: one for holding base information about the
> document ( author, createtime , ...) , one table for numeric fields, another
> for their contents, one for string-content and so on. This ends up in loads
> of related tables and I' worried about performance problems running queries
> on this structure. ( concerning an amount of estimated 100000 up to 200000
> documents. )
>
> So my question(s) is/are :
> (1) Is their a more simple or efficient way to implement such datastructures
> on a relational database ?
> (2) Or should I implement a database system on my own ? :-((
> (3) Concerning relational databases : Would Oracle 8 be the best decision to
> avoid such performance problems ?
>
> Any help appreciated,
> thx
> Charly

        No one, except a salesman, can positively state that any RDBMS product is the best decision. Oracle 8 can definitely do the job for you.
You can store the documents in the database as blobs.

        Based on your discussion I believe that you want to create a set of indexes that will allow the rapid identification and retrieval of the appropriate documents.

        An approach I am familiar with builds keyword lists for every document. A query that acts as a search engine is then written and the users interact with the "Search Engine" to identify the desired documents.
Depending on how sophisticated your "Search Engine" is either it or the user then extracts the desired document.

-- 
Jerry Gitomer
Once I learned how to spell DBA, I became one.
Received on Fri Feb 04 2000 - 00:00:00 CET

Original text of this message