This can provide better scalability by allowing a data set to be fully cached even if it is too large to fit in the memory of any single system. It can also provide better write performance by spreading entries into multiple independent replication sets.

Entry balancing breaks the data into backend sets. Each backend set has its own proxying request processor, which has its own load-balancing configuration which might optionally include criteria-based load balancing. Each of these load-balancing algorithms can be configured to use automatic backend set discovery in exactly the same way as in a non-entry-balanced configuration. The backend sets must be explicitly configured, but the servers within each set can be automatically determined from the information in the topology registry.