[OT] Re: Loadbalancing Datbases - mysql

[Nigel Hamilton <nigel@xxxxxxxxxxx>]

> Now it looks that there will be a solution for that. I found a thread in the
> myslq-cluster mailing list and a link in the manual. Mysql 5.1 will be able
> to handle disk-based records in the cluster (see link and thread) ... and I
> think this will be a very nice solution (furthermore the mysql cluster code
> is more stable and more tested).
>
> Thread:
> http://forums.mysql.com/read.php?25,20801,20801#msg-20801
>
> Link:
> http://dev.mysql.com/doc/mysql/en/mysql-5-1-cluster-roadmap.html

Aarrgh ... it always seems the MySQL feature I need is always N+1 versions 
away! :-)
Certainly the RAM requirements of the initial MySQL Cluster design makes 
it unworkable for me so a hybrid RAM + disk-based system is welcome news.
We have a slightly unusual situation in that our writes may equal or 
exceed reads - I was planning on using "memcached" to help with handing 
off as many reads as possible to a distributed RAM bucket. But I'm still 
thinking of ways to distribute writes - ideally at the application level 
all I need is a database handle that connects to one big high speed 
amorphous distributed DB.
My current short term solution to distributing writes is to use a MySQL 
heap/ram table on each node which acts as a bucket collecting writes which 
is periodically emptied to the Master's disk.
But I'm not sure how far I can get with this approach. I'd be interested 
in hearing how others have made their MySQL-based application scale up?
Nige