The way memcached works across multiple servers is that it distributes the data across the memcached instances running on all the servers. The client determines which memcached instance to connect to when getting a particular item.
It all works fine and I really like it.
BUT For a particular dataset, I want the data available super quickly. I do not want to make the trip to another server's memcached to get the data. Essentially I want a replicating memcached - i.e. the full data set is held on each server's memcached instance.
I want to do this for performance (not for availability, redundancy etc.)
I see something called repcache - would this be the right way to go? are there other solutions? Is there a flaw to my thinking?
Just wanted to get a bit of feed back before I strike out in any particular direction.