November 2018 Summaries

CockroachDB 2.1 demonstrates significant enhancements in scalability and cost-efficiency compared to Amazon Aurora, achieving 50 times the scalability at less than 2% of the price per transaction per minute (tpmC). Unlike Aurora, which compromises on isolation levels for performance, CockroachDB maintains serializable isolation, safeguarding data from fraud and loss while supporting large-scale operations. Through the TPC-C benchmark, CockroachDB 2.1 achieved a remarkable 631K tpmC and improved its efficiency by reducing the number of nodes required for operations, resulting in substantial cost savings. Additionally, the KV benchmark showed linear throughput increases with additional nodes, while the 3-node cluster improvements translate to more throughput with fewer resources, benefiting deployments that don't require global scale. CockroachDB's enhancements make it a cost-effective and resilient database solution for various customer needs.

As of March 2019, CockroachDB announced the capability to import MySQL database dump files, facilitating a smoother transition for users migrating from MySQL to CockroachDB, a globally-distributed SQL database. This feature, though initially in beta, allows data from MySQL to be imported with a single command, streamlining the previously cumbersome process of manually specifying schemas due to the limitations of CSV-based imports. The integration involves using an adapted MySQL parser from the Vitess project to translate MySQL dialects into CockroachDB's system, ensuring compatibility with its data types and efficient data distribution across nodes. Despite some known issues, such as difficulties with full-text indexes and certain invalid date values, CockroachDB is confident that this new import functionality will significantly ease the transition for users, evidenced by its successful application in large-scale migrations like Baidu's. Users are encouraged to follow the MySQL migration guide and report any bugs encountered during the process.

Cockroach Labs has introduced a new cost-based SQL optimizer in its 2.1 release to enhance the performance and scalability of CockroachDB, particularly in handling complex queries. Unlike heuristic optimizers, which rely on preset rules that can sometimes fail to find the most efficient query execution plan, the cost-based optimizer evaluates multiple execution plans by assigning a cost to each, selecting the one that minimizes resource usage and execution time. This approach allows for more accurate and flexible query optimization, leveraging detailed database statistics to make informed decisions. The development of this optimizer involved a focused team effort and educational initiatives to understand state-of-the-art optimization techniques. The new optimizer supports features such as correlated subqueries and the automatic planning of lookup joins, and it employs a data structure called a memo to efficiently manage the numerous potential execution plans for complex queries. This structure helps in storing and evaluating plans without excessive memory use by recording only unique plan components. Cockroach Labs plans to continue refining these optimizations and invites community contributions and potential recruits to join their efforts in advancing distributed SQL engine technology.

CockroachDB 2.1 introduces significant enhancements aimed at improving scalability, ease of migration, and operational efficiency for global applications. The new version offers 5x scalability improvements for transactional workloads, enabling developers to handle more complex and growing data demands without sacrificing data consistency. The release facilitates easier migration from MySQL and Postgres through enhanced import capabilities, supporting developers transitioning from these popular databases. Additionally, CockroachDB 2.1 introduces a managed service to simplify the deployment of low-latency, multi-region clusters and provides tools for better performance monitoring and debugging. The update also includes beta support for Change Data Capture (CDC) to stream changes to systems like Apache Kafka, promoting real-time data integration and event triggering. Managed CockroachDB, currently in limited availability, promises to streamline the deployment of geo-distributed clusters by handling networking and provisioning tasks, allowing teams to concentrate on application development.