NetworkTopologyStrategy

NetworkTopologyStrategy is a replication strategy used by Apache Cassandra for distributing data across multiple data centers (and racks). It allows per-data-center replication factors, enabling replicas to be allocated independently in each data center, rather than applying a single global factor.

Usage: When creating a keyspace, specify options like {'class': 'NetworkTopologyStrategy', 'DC1': '3', 'DC2': '2'}, where DC1 and DC2 are the names of data centers as detected by the configured snitch. The strategy uses a topology-aware placement algorithm that seeks to place replicas across distinct racks within each data center, improving fault tolerance against data-center failures and rack failures, and supporting locality for latency-sensitive clients.

Requirements: A network topology-aware snitch (for example GossipingPropertyFileSnitch) must be configured so the system can identify data centers and racks. The NetworkTopologyStrategy works with the cluster's token distribution and vnodes to place replicas within each DC according to the specified replication factors.

Notes: You can only specify replication factors for data centers that exist in the cluster; if you include a DC with a factor of zero, that DC will not receive replicas. Changing replication factors or adding data centers requires a repair to ensure consistency across replicas.

Comparison: Unlike SimpleStrategy, which uses a single replication factor across all nodes, NetworkTopologyStrategy distributes replicas per data center. It is generally recommended for production clusters spanning multiple data centers.

See also: Apache Cassandra, Replication, Data center, Snitch.