By default, distributed maps have one replica to insure failover; when an entry is added or updated in the map, Hazelcast synchronously creates a replica on another node.
Hazelcast manages the replicas with a backup operation; it provides the following two kinds of operations:
Sync backup operations
Async backup operations
Cellar only uses a sync backup operation to ensure that the replica is complete before it moves forward. You can increase the number of replicas using the backup-count element in the etc/hazelcast.xml file.
By default, the replicas are used only when a node goes down; Hazelcast can reload the data located on the failed node using the replica present on another node.
However, in a real runtime environment, a node has to call the node that stores the data, even if it holds a replica of this data locally.
To increase the read performance, you can allow the nodes to read their local replicas instead of actually calling the nodes that store the data. To do this, you...