On Tuesday 23 January 2001 13:35, Mauricio Breternitz wrote:
>Folks:
>
> I'm attempting to develop a parallel MySql on my company's
>system (see http://www.timesn.com )
>my initial goal is to enable inter-query parallelism onto a shared
>file system).
>
>The (simple) idea is to capture sequences of queries and ship each query in
>a rond-robin (or load-balancing) fashion to the available processors.
>
>The approach I'm taking is to run one copy of the server
>per processor & extend the file locks such that they become
>global locks to prevent multiple processors from stepping on each other.
>
>However, I do not want to introduce excessive serialization (e.g, making
>every call to pthread_mutex_.. into a global lock.
>
>Can folks more familiar with the source tree than I am (I got the
>tree from bitkeeper) point me to places where I should lock file
>access ? (I've been chasing calls to my_open to see where I would
>place locks aroung file open / file close cals). Of course I'm open
>(no pun intended...) to better suggestions
Your best bet is to use replication for load balancing, or just beef up the
server on CPU power as much as you can. Paralelizing one query is not a very
good idea, I would actually go further and say that you will not be able to
do it without re-writing MySQL from scratch.
--
MySQL Development Team
__ ___ ___ ____ __
/ |/ /_ __/ __/ __ \/ / Sasha Pachev <sasha@stripped>
/ /|_/ / // /\ \/ /_/ / /__ MySQL AB, http://www.mysql.com/
/_/ /_/\_, /___/\___\_\___/ Provo, Utah, USA
<___/
