no matter what size is your application distributable should be always under consideration. otherwise it becomes a full pain full reimplementation of whole application. yes reimplementation is always good but for developers it is always painful and for customers it is always very but very expensive.

here is some key points

Model extends Objects implements java.io.Serializable

Your model objects and communication objects should be Serializable. Do not forget serialization is the most important feature. it necessary for caching,messaging,storing,…

API replacements

Every instance should be aware of its copies in cluster. so you may need some api replacements like

Map->DistributedHashtable : nice seamless implementation from JGroups.

Cache->Distributable cache: (eg:EhCache,OSCache)

Observer,Listener>JMS Queues : for whole notifications in the cluster.

Timer->Quartz distributable scheduler

Related posts:

1. Clustering Acegi via JGroups (DistributedHashtable)
2. JGroups-Spring In Action
3. DistributableEventPublisher for Spring via JGroups
4. hibernate native Id generator tip
5. multiple tomcat instance – single setup