AnalyticDB for MySQL builds a distributed, strongly consistent, highly available, and highly reliable storage system in compliance with the Raft protocol at the storage layer. This system provides higher availability of reads and writes than the traditional relational database management system (RDBMS) that uses the primary/secondary protocol as well as the commonly used distributed middleware system that uses the database sharding and table partitioning mechanism.

High availability of reads and writes
AnalyticDB for MySQL uses Multi-Raft to manage data shards. Each data shard consists of three replicas. A leader is selected, logs are replicated, and logs are applied to state machines in each shard in compliance with the Raft protocol. To ensure high-availability reads and writes, at least two of the three replicas must be running. This allows the node where one replica resides to shut down or restart. Even if one of the replicas fails, the databases can still support read and write operations.