[thelist] Relational DB Help

Andrew Forsberg andrew at thepander.co.nz
Sat Feb 23 15:01:01 CST 2002


Hi Dave

>Also, lets say we're using MySQL and not a super-scalable, massively
>relational database like Oracle with great indexing functionality... MySQL
>would choke.

Re: the Tim Perdue benchmarks which suggest MySQL chokes under heavy
load, the MySQL docs have been updated and question those tests'
conditions:

http://www.mysql.com/doc/M/y/MySQL-PostgreSQL_benchmarks.html

It appears that a lot of the speed, and stability under load,
problems were related to a sub-production quality build (presumably
by Tim Perdue), and an older linux system with a poor thread
implementation. Both of which are discouraged strongly in the build
docs for those same reasons.

>What does he gain?  The ability to use specialized tools for searching, a
>less bloated database, and his page creation won't be bogged down by someone
>running a query for "the" in a potentially huge database because they would
>be separate process/applications.

Even if MySQL's full-text search functions didn't exclude words under
4 letters, or words which are exceedingly common in over 50% of the
indexed field(s) of the table -- and this is more or less how MySQL,
by default, deals with full-text searches for common words like 'the'
or the name of the site -- it's a fairly trivial task to disregard
short words from searches before they even get to the db.

>Just thoughts... I might seem to be advocating text files... but normally I
>wouldn't (it's not a good practice)... but it might make sense here unless
>he's using a big relational database on a real server system.

The only way sure way to tell would be to test it, IMO.

Cheers
Andrew

--
Andrew Forsberg
---
uberNET - http://uber.net.nz/
the pander - http://thepander.co.nz/



More information about the thelist mailing list