Handling portal caches across cluster websphere portal

Cache :- Portal has built in cache mechanism like dynamic cache , distributed map,distributedCacheObject . Where you can share objects which saves time by retrieving it from cache instead of database call. If client has four nodes under a cluster how does then cache will be shared across the nodes. This should be fundamental point any web-sphere administration should never forget.To avoid this portal has data replication system which helps in replicating the data between all nodes.

Data replication service (DRS) is an internal WebSphere Application Server component that replicates data. Transport for sending data or objects or events from one managed server to another.This take care of syncing data between two nodes.

• Uses HAManager and DCS data stack frameworks to accomplish replication. 
• Has the notion of replicas; however DynaCache only works with Entire Domain.
• Total Number of replicas = Number of servers in the replication domain -1 
• Responsible for serialization and de-serialization of messages. 
• DRS Bootstrap will be expensive, if aggressive replication occurs during startup.
• DynaCache, HTTPSession, Stateful Session Beans and SIP are major consumers. 
• Data transfer channel can be encrypted. In practice no-one does this. 
• DCS uses a star topology for synchrony resulting in scalability bottlenecks

WAS console :- Navigate to web-sphere application server console -> Resources -> object cache instances -> click on cache instance . Then select enable data replication then select replication type based on your requirement. Better suggested in Not_Shared if required use push or push_pull but don't go with pull option its very expensive.

In all of the following modes, the invalidation's will be sent across servers for data consistency. :-
 PUSH mode :- Both cacheId and CahceData will be sent to all servers in cluster
 PULL mode (Not recommended) :- Server requests the data from other servers in cluster, when not present in its cache
 PUSH PULL mode:- Only cacheId will be sent to all servers in cluster. When a server need an entry that is in PushPull table, it requests the server that has copy.
Not Shared :-In this mode the cache data will not be shared across servers. But the invalidation events are sent to all servers.

Source :- 
http://www-01.ibm.com/support/docview.wss?uid=swg27017355&aid=1