Leaving the preferred.node.indices property blank in all cluster-configuration files results in a basic deployment case. An advantage of this approach is simplicity, including the option of using straightforward load-balancing strategies such as round robin. A disadvantage is that as additional nodes are added, the throughput improvement rate that clustering offers may decline as the state-replication overhead increases.

The following diagram illustrates this node-sharing approach. Requests in this deployment are directed to all nodes.

Runtime state management architecture - Sharing all nodes