| Mode | How It Works | Pros | Cons |
|---|---|---|---|
| Synchronous (strong) | Primary waits for replica ACK before commit. | Zero data loss (RPO = 0) | Higher latency; limited distance |
| Asynchronous (eventual) | Primary returns immediately; replicas catch up. | Fast writes, long‑distance replicas | Possible data loss on primary crash |
| Multi‑primary / Leaderless | Any node can accept writes; conflicts resolved later (e.g., last‑write‑wins, CRDT). | High write availability | Conflict resolution complexity |