Tuning FULLTEXT performance (was Re: Memory upgrade and swap partition size)

Fred puissante at lrc.puissante.com
Wed Sep 27 09:20:01 EDT 2006


On Tuesday 26 September 2006 19:31, Drew Van Zandt uttered thusly:
> Master --> slave with all inserts on the master copy?  Lets you at
> least split the load onto 2 servers...

That's an idea. Though I may have to do it the other way, and just have 
replication going in both directions (I already have replication going in 
one direction).

Still, the number of updates coming in will themselves lock up the table, but 
not as quickly.

I think your idea coupled with batching the updates for "end of day" could 
work.

What I really need is for FULLTEXT to be implemented for InnoDB tables. THAT 
would be sweet. MyISAM tables stinks when it comes to heavy updates, though 
are wicked fast on the queries. But even there under certain circumstances 
the InnoDB tabes are faster.

My big headache is that I have a HUGH database that is queried in a sparse 
fashion across then entire set. The normal gains one would get from caching 
is simply out. Tuning for performance under these circumstances is a 
challenge, to say the least. Though, I did get everything working nicely 
until I introduced the FULLTEXT stuff. :-(

-Fred



More information about the gnhlug-discuss mailing list