On Thu, 2006-01-12 at 15:39 +0100, Paul J Stevens wrote:
> Simon Gray wrote:
> 
> > http://www.danga.com/memcached/
> 
> Which is very cool stuff. Definitely a planned thing.
> 

It is, Sean Chittenden of pgmemcache has been pushing it for over a
year. When we're ready for multimaster replication, using the token
passing algorithm that was laid out in the list (sorry, can't find it in
the archive right now), it will be a good time to get into memcached.

One of the issues with memcached is that you need to have your cluster
membership figured out, and/or, a good system for instances of dbmail to
use in case they cannot contact a memcached process or if that node
somehow managed not to be a member of the memcached cluster. When we
figure out DBMail clustering, these questions will also come up, so we
can solve them in parallel.

Aaron

Reply via email to