Read/Write Separation
마지막 업데이트 시간:2026-02-04 16:13:23
Overview
Read/write separation is a performance scaling solution for read-intensive, write-light workloads. By distributing read requests to replica nodes, it alleviates pressure from concentrated reads on hotspot data. It supports up to one master node and five replica nodes, delivering up to 5x read performance scaling capability.
Architecture Principles
Read/write separation is implemented through a built-in proxy component for automatic routing and distribution of requests. Client requests are uniformly sent to the proxy, which automatically identifies the request type, routes write requests to the master node for processing, and distributes read requests to replica nodes according to the load balancing policy. After read/write separation is enabled, the master node only handles write requests, while read requests are handled by replica nodes, thus achieving linear scaling of read performance.
Supported Versions
Read/write separation is compatible with Redis 4.0, 5.0, 6.2, and 7.0 as well as Valkey 8.0. It applies to both standard and cluster architectures, and can be enabled when the number of replicas is greater than or equal to 1.
Enabling Methods
Console: On the Node Management page, you can enable the read/write separation feature. For specific operations, see Read-Only Replica.
TencentCloud API: You can enable read/write separation through the API EnableReplicaReadonly and disable read/write separation through the API DisableReplicaReadonly.
피드백