EnterpriseDB Corp., which sells a commercial version of the popular open-source PostgreSQL database management system, today said it enhanced the hybrid features of its EDB Postgres AI platform with unified management and features that speed artificial intelligence development and processing.

Citing its own research that found that 56% of enterprises use a hybrid model combining public and private cloud resources for mission-critical workloads, EDB said the new features enable rapid application deployment and management across mixed infrastructure with a single control point.

EDB announced the AI version of PostgreSQL in the spring, adding the capability for users to conduct analytics on transactional data through a lakehouse that runs adjacent to the production database. A lakehouse is a modern data architecture designed to provide the scalability, flexibility and cost-effectiveness of a data lake with the data management and performance capabilities of a data warehouse.

“Many of the things we announced in spring were only available in our cloud-hosted system, so if you wanted to take advantage of them you had to move to the cloud,” said Jozef de Vries, the company’s chief product engineering officer. “We’re now taking our cloud-managed experience and providing it as packaged software with connectivity between the two.”

A control plane based in a customer’s data center can manage both on-premises and cloud databases. Unified management is currently available only on the company’s hosted database service but will be extended to other cloud providers in the future, de Vries said.

AI acceleration

The new release also includes what the company calls an AI Accelerator based on EDB’s Pipelines Extension and preloaded with pgvector, an open-source PostgreSQL extension that enables the storage, indexing and querying of high-dimensional vector data. The combination allows users to test and deploy generative AI applications such as chatbots and recommendation engines by automating the entire data pipeline – including embedding generation, storage, indexing and retrieval — within the PostgreSQL environment.

“In a traditional Postgres plus pgvector setup, you have the ability to store vectorized embeddings indexed, but then when you do data retrieval or data capture, you have to reach outside of your Postgres ecosystem,” de Vries said. “We’re building out the capabilities within our AI accelerator extensions so you can do data capture and embedding generation, store those vector embeddings, index them and facilitate the retrieval all through the P-SQL interface. Developers don’t have to leave they environment that they’ve been accustomed to working with.”

Procedural SQL is an extension of Structured Query Language that allows for procedural programming features and complex logic to be executed within a database.

New multimodel data management features provide near real-time online analytical processing capabilities with 18 times greater cost efficiency and 30 times faster performance than baseline PostgreSQL, the company claimed.

“Multimodel data management emphasizes the different data types that Postgres inherently is able to support,” including structured data, relational data, JavaScript Object Notation and time-series data notation, de Vries said. “The emphasis is on bringing those data types more to the forefront of the Postgres experience and shifting away from having multitudes of specialty databases for all the different data types.”

An Analytics Accelerator allows users to query columnar data in external object storage using SQL. It also supports tiered tables functions, which uses logical replication to move in frequently accessed data into another storage tier on a cluster without sacrificing performance.

Image: SiliconANGLE/DALL-E