Memcached Clusters?

Reinis Rozitis roze at
Tue Oct 24 07:42:44 UTC 2006

> It would be good if the servers all shared the same data (or some other
> sort of redundancy). This would mean that if one of the servers went
> down, the other servers would still retain the data and as such it would
> provide high-availability for the system.

Well on the other hand if you split the data between a pool of servers there 
is less amount on each so it works faster this way.

You could just make client side code such way - do not add the servers to 
pool (dont use add_server()) but just to do set() command on each of 
Memcaches and then do a random get() (if the first connect fails (eg server 
went down) connect to the second server)).

Practicaly you can implement pretty everything on client side code..


More information about the memcached mailing list