Dynamic Cluster

NCache provides high availability and hot scalability through its peer-to-peer Dynamic Clustering architecture. Nodes continuously monitor cluster health using an overhead-free heartbeat mechanism. When servers join or leave, the coordinator updates cluster membership and automatically triggers Dynamic State Transfer, redistributing data in the background to maintain redundancy and balanced storage without disrupting client applications.

State transfer behavior depends on the cache topology. Replicated Caches synchronize the full dataset with new members, while Partitioned and Partition-Replica Caches move only the affected data buckets, minimizing network traffic. Partition-Replica also ensures zero downtime by instantly promoting replica buckets if a server fails.

For planned maintenance and network disruptions, NCache provides Maintenance Mode and Split-Brain Recovery. Maintenance Mode pauses state transfer while a server is temporarily offline, avoiding unnecessary data movement. If the cluster becomes partitioned, NCache automatically detects the split and safely reunifies the cluster using majority-node consensus once connectivity is restored.

In This Section

Dynamic Cluster
A peer-to-peer architecture that automatically updates membership and allows on-the-fly server scaling without application downtime.

Dynamic State Transfer
An automated background process that reshuffles and balances data across nodes whenever cluster membership changes.

Split-Brain
A self-healing safeguard that detects network partitioning and automatically merges isolated sub-clusters once connectivity is restored.