Terracotta

Open Source Clustering For Java

Architecture

JVM Level Clustering

Heap-Level Replication: Selectively share object graphs across the cluster
Large Virtual Heap: manage heaps that are much bigger than what can be contained in a single JVM.
Cluster Wide Locking Semantics: distributed wait/notify and synchronized
Runtime Monitoring and Control: View in real-time objects in the heap
Fine-Grained Changes: Terracotta does not use serialization, and therefore can push field-level changes to the cluster. This translates into better network efficiency for replication.
No Serialization: Because Terracotta clusters at the JVM level, no serialization code is required.