Postgres Without Limits

PostgreSQL’s built-in cost-based optimizer (CBO) is critical for handling multiple and diverse workloads. YugabyteDB’s new Adaptive CBO extends the range of PostgreSQL’s CBO for high-scale and multi-region applications. It determines an optimal query plan that considers whether data is co-located, automatically sharded, or even distributed across zones or regions. It also implements core Postgres capabilities such as extended table statistics, parallel queries, and bitmap scans in a fully distributed system.

The Adaptive CBO capability is built on several core features that allow us to embrace and extend PostgreSQL’s CBO. These features, several of which are available in Postgres, have been adapted or even reimagined for a distributed system.

YugabyteDB’s Adaptive CBO picks the optimal query plan in 93% of TAQO Framework tests, which is close to the 97% score achieved by Postgres. The TAQO Framework is a query optimizer testing framework for Postgres compatible databases based on research at the University of California at Davis.

Distributed SQL databases typically federate data and query processing across nodes. While this enables workloads to scale, it can degrade the performance of applications built for traditional databases. In this case, developers are forced to choose between lower latency and more scalability.

YugabyteDB eliminates this forced choice by automatically determining whether to store tables together in a colocated manner for lower latency, or shard and distribute them to achieve massive scale.

Smart Data Distribution enables YugabyteDB to support over 50,000 database objects per DB – tables, indexes, sequences, and more – compared to 2,000 without smart data distribution.