How memcached chooses from group of servers
Memcached is a ditributed caching solution for Java enterprise applications. It has a client as well as server. Server needs to be installed and can be part of multiple nodes.
If you have more than one Memcached server, the servers won’t know about each other. Instead, you’ll configure your Memcached client so that it knows all the available Memcached servers. For example, if your application creates a Java object on AppServer1 and calls the set() method of Memcached, then the Memcached client will figure out which Memcached server that entry goes to. It will then start communicating with that Memcached server only.
In its default configuration, the Memcached client uses very simple logic to select the server for a get or set operation. When you make a get() or set() call, the client takes the cache key and call its hashCode() method to get an integer such as 11. It then takes that number and divides it by number of available Memcached servers, say two. It then takes the value of the remainder, which is 1 in this case. The cache entry will go to Memcached server 1. This simple algorithm ensures that the Memcached client on each of your application servers always chooses the same server for a given cache key.
On the down side there’s no concept of replication. This means if one Memcached node goes down, you lose the keys on that node. If you’re only connecting to a single node then you lose all Memcached.