What it is: An asynchronous copy of a database used primarily for scaling reads.
What it’s for:
Offload read traffic (reporting, dashboards, read-heavy workloads).
Improve performance for read operations.
Key ideas:
Replication is asynchronous (replica may lag behind).
Read replicas have their own endpoint.
Some engines allow promotion to standalone DB (disaster recovery pattern, but not “automatic same endpoint” like Multi-AZ).
Exam cues:
“read-heavy workload, scale reads” → Read Replica.
“replica lag is acceptable” → asynchronous replica.
Hard words:
*asynchronous* /eɪˈsɪŋkrənəs/: bất đồng bộ
*lag* /læɡ/: độ trễ (chậm hơn dữ liệu gốc)
*promote* /prəˈmoʊt/: nâng cấp (thành primary)