System architecture

Updated at:

The architecture has two data centers (DCs). DC1 and DC2 each contain four nodes.

Nodes

  • A Cassandra instance consists of one to three DCs. Each DC contains 1 to 500 nodes. Each node runs one Cassandra process.
  • Cassandra has no master or proxy nodes. All nodes are peers. The system selects two nodes in each DC to act as seed nodes. These seed nodes do not require a separate process and impose minimal load on the system.
  • Clients connect to each server-side node for high performance.
  • Different data centers can have different numbers of replicas. This configuration supports cross-data center disaster recovery, backups, and batch analytics.

Nodes

Each node provides 160 GB to 16 TB of storage space. A 500-node cluster can provide a maximum of 8,000 TB, which is approximately 8 PB.