To ensure data remains available after a single availability zone failure, the active-active cluster uses replication mechanism instead of erasure coding (EC). While EC's 1+1, 2+2, 3+3, and 4+4 peer modes ensure data integrity, their space utilization is limited. Additionally, cross-domain data computation during anomaly recovery can significantly reduce performance. Therefore, in the active-active cluster mode, ABS only supports the replication policy, limited to 3 replicas, ensuring that after a replica fails, data degradation will not be triggered by cross-domain traffic. Replica allocation follows the principle of reserving 2 replicas in the primary availability zone and 1 replica in the secondary availability zone. Adjustments are made in specific scenarios.
