IDEA: Hierarchy of caches for high performance AND high capacity.

Kevin Burton burton at
Wed Nov 1 00:49:26 UTC 2006

On 10/31/06, Ask Bjørn Hansen <ask at> wrote:
> On Oct 31, 2006, at 15:08, Kevin Burton wrote:
> > MySQL doesn't scale ;)
> I think you mean "vertical scaling sucks".  MySQL can be used
> horizontally, as you know.  :-)   (sometimes replication, sometimes
> partitioning, sometimes a combination).

Yes.... but that sucks as well because you need to code around your

All these tricks to make MySQL scale are hacks and they really shouldn't
make it up into application space.

I think as an industry we've had a major cop out and come up with excuses
not to build a real cluster scale database.

The thing is that a decent clustering database would scale from small
installs to HUGE installs without any hacks like buying big boxes, cluster
partitioning, any of the other tricks people do to MySQL to squeeze
performance out of a bad idea.

> Especially replication.........
> I meant using MySQL as a "cache store" -- you don't want to replicate
> that.

oh ...... yeah.... I'm not suggesting that at all :)

MySQL (non-cluster[1]) can work beautifully as a substitute for
> memcached when you have more data to store than you have memory - and
> you can afford waiting for disks.

 Maybe....... if you do hash routing..... of course MySQL won't be able to
handle anywhere near as many connections as memcached...

Tulegacache would do a better job at memcached either way....


Location: San Francisco, CA
AIM/YIM: sfburtonator
Skype: burtonator
Cell: 415-637-8078
-------------- next part --------------
An HTML attachment was scrubbed...

More information about the memcached mailing list