Inside RadonDB: How QingCloud’s MySQL‑Based NewSQL Database Tackles Distributed Challenges

Open‑source release

RadonDB is a next‑generation distributed relational database built on MySQL and released under an Apache‑compatible license. The source code, issue tracker, and pull‑request workflow are hosted on GitHub at https://github.com/radondb. Documentation and binary releases are available from the project website https://radondb.io.

Design goals

The primary goal is to provide a fully MySQL‑compatible front‑end so that existing applications and DBA tools can be used without code changes or additional learning. By preserving the MySQL protocol and SQL dialect, data migration from single‑node MySQL to RadonDB is seamless.

Architecture overview

RadonDB consists of two tightly coupled components:

Radon – the SQL layer that parses client queries, routes them, and coordinates distributed transactions.

Xenon – the storage layer that embeds MySQL instances and adds NewSQL‑style high‑availability using the Raft consensus algorithm.

Key architectural features include:

Raft‑based leader election integrated with MySQL so that when the primary node fails, Raft automatically promotes a new leader.

Parallel replication from MySQL’s native binlog mechanism enables rapid replay of pending writes after a failover.

Automatic sharding – sharding rules are defined once and the system automatically migrates data to the appropriate shards.

Elastic scaling – nodes can be added or removed; data is rebalanced safely without manual intervention.

Distributed transactions – the SQL layer coordinates two‑phase commit across shards, leveraging MySQL’s ACID guarantees.

High‑availability performance

In a live demo, 64 concurrent writer threads were active when the primary node was killed. Raft elected a new leader and MySQL’s parallel replication restored service in approximately three seconds, demonstrating near‑instant recovery under load.

Automatic sharding and scaling

Traditional distributed databases require DBAs to manually configure shard maps, validate data placement, and rehearse migrations. RadonDB automates both the definition of sharding rules and the data movement, eliminating error‑prone manual steps. Scaling operations (adding or removing nodes) trigger an automatic, consistent rebalancing process that preserves data integrity.

Distributed transaction support

RadonDB currently implements distributed transactions across shards using a coordinator in the Radon layer. Future work aims to expose more of MySQL’s native optimizations (e.g., GTID, group replication) to improve throughput and latency.

Tooling

The high‑availability helper MySQL Plus is bundled with RadonDB and released as a separate open‑source project. It can be downloaded from the official RadonDB download page linked from the GitHub repository.

Competitive advantages

Compared with other Chinese open‑source distributed databases such as OceanBase and TiDB, RadonDB’s MySQL‑centric architecture offers several benefits:

Leverages the mature MySQL ecosystem (performance patches, security updates, community tooling) which reduces development and maintenance cost.

Provides a lower entry barrier for teams already familiar with MySQL, avoiding the steep learning curve of completely custom storage engines.

Combines NewSQL concepts (Raft consensus, distributed transactions) with MySQL’s proven storage reliability, delivering both robustness and ease of adoption.

Images