MySQL Cluster

MySQL Cluster (NDB Cluster) is Oracle's distributed database designed for applications that require linear scalability and 99.999% availability. It targets real-time, high-throughput workloads such as telecommunications, online gaming, and financial services where downtime is unacceptable. The cluster stores data primarily in memory for low-latency access while maintaining transactional consistency across partitioned datasets. It is best suited for organizations that need a single database to handle both relational and NoSQL access patterns, with active-active geographical replication across multiple sites. MySQL Cluster is not a general-purpose replacement for standalone MySQL; it is a specialized product for users who must scale writes horizontally and survive node failures without manual intervention.

MySQL Cluster achieves high availability by maintaining up to four synchronous replicas of every dataset, ensuring no single point of failure. A single cluster can store over 100 TB of data and scale to hundreds of nodes. It uses a shared-nothing architecture where data is automatically partitioned across nodes, and each node is aware of data locality, minimizing network hops. The parallel distributed query engine allows complex queries to be executed across partitions concurrently. Online scaling and schema upgrades are supported without taking the cluster offline, which is critical for 24/7 operations. The cluster also supports multi-site active-active replication, enabling disaster recovery across geographically separated data centers.

In the MySQL clustering market, MySQL Cluster competes directly with Amazon Aurora, Galera Cluster, InnoDB Cluster, and third-party tools like ClusterControl. Amazon Aurora offers similar high availability but is limited to a single writer node and does not support the same level of horizontal write scaling. Galera Cluster provides multi-master replication but lacks the integrated distributed query engine and data locality features of NDB Cluster. InnoDB Cluster is simpler to set up but does not match NDB Cluster's ability to handle over 100 TB or provide 99.999% availability. MySQL Cluster is also priced per server socket, which can be cost-effective for high-core-count servers compared to per-core licensing from competitors like Microsoft SQL Server or Oracle Database.

The honest trade-offs with MySQL Cluster are significant. It lacks advanced features like full-text search and partial indexes, which limits its use for content-heavy applications. Commands can occasionally be buggy, and server response times may be slower than single-node MySQL for simple queries due to distributed coordination overhead. Historically, it has weaker support for full ACID compliance and sophisticated query optimization compared to PostgreSQL or Oracle Database. Scaling up with complex queries and very large databases can be challenging, and additional backup services are often required because native backup tools are limited. Organizations must weigh the need for extreme availability and horizontal scalability against these operational and feature constraints.

Priced per server socket via annual subscription: Cluster CGE (1-4 sockets) at $10,700/server/year; Cluster CGE (5+ sockets) at $21,400/server/year. Community Edition is free under GPL.