Distributed Redis Architecture
The distributed Redis adpots a Proxy + Redis sharding architecture. The proxy supports only standalone Redis, while Redis shards in the cluster can be used without a proxy, but only in a native cluster mode. The usage method should be selected according to the type of client (for command limits, refer to the FAQs). Redis sharding is based on the master-replica version of the Redis resource pool, effectively breaking through the single-threaded bottleneck inherent in Redis. It supports online expansion, which can significantly meet the business requirements for large capacity or high performance in Redis.
1. Distributed Redis Architecture
Each Redis shard in the distributed Redis setup is a master-replica version of Redis that shares a resource pool with the master-replica version of Redis. Its architecture is identical to the high-availability architecture of the master-replica version of Redis. The capacity limitations and the rules for expansion and contraction of a single shard are consistent with those of the master-replica version of Redis. The proxy for distributed Redis is designed with high availability in mind. The container-based proxy will be automatically migrated to other hosts by the system in the event of a host failure. The Kuaijie-based proxy consists of one master and one backup host, and the high-availability IP will be migrated to the backup host by the system after the master fails.
2. Configuration Mode of Distributed Redis Proxy
The master-replica type proxy offers an IP gateway for each proxy, and users can freely configure the proxy gateway.
Under the load-balancing proxy mode, the cluster provides only one IP gateway. Business requests are sent to ULB4, and ULB4 Server performs failed-down and request balancing.
3. Internal Architecture of Distributed Redis Proxy
The internal architecture of Distributed Redis Proxy is composed of multiple Proxy processes (Redis Proxy). The system kernel balances client connections and binds them to a specific Proxy process. Within the Distributed Redis Proxy, there is one process (BusyProxy) that handles multi-key commands to process complex queries.
The Distributed Redis cluster employs a multi-Redis node architecture, where each single node is based on the master-replica version of the Redis resource pool. Backup nodes maintain high availability, allowing the system to easily overcome the single-threaded bottleneck of Redis itself. This architecture supports online expansion, which can significantly meet the business needs for large capacity or high performance in Redis..
