Key mechanisms were initially implemented in the Tantor XData Gen3 database engine.

Tantor Labs announced the release of a new version of the Tantor Postgres 18.3.0 database management system, built on PostgreSQL 18. In this “major update,” developers combined the capabilities of the main PostgreSQL 18 and 18.3 branches with their own technological solutions.

One of the important innovations is the implementation of the Commit Sequence Number (CSN) mechanism. This technology, previously used in the high-performance Tantor XData Gen3 database engine, optimizes the determination of row visibility in snapshots. It addresses the scalability problem of the MVCC architecture, where obtaining a snapshot required locking and checking all active connections, creating a bottleneck under high concurrency. Now, this process is independent of the number of connections, which significantly increases system throughput.

To optimize data storage in the DBMS, the Compression Storage Manager (CSM) has been introduced. The new mechanism provides transparent compression of entire data pages when writing to disk, using pglz, lz4, and zstd algorithms. Compressed pages are stored as blocks of a smaller, configurable size (1024, 2048, or 4096 bytes), while maintaining the standard PostgreSQL structure.

Data lifecycle management is handled by the new pg_ilm (Intelligent Lifecycle Management) extension. The system automatically moves information between different storage modes, analyzing its “state”: access frequency, format (row or columnar), and placement in the storage layer. This allows offloading fast disks by migrating less demanded data and maintaining stable performance without manual intervention.

In the field of security, auditing and anonymization capabilities have been expanded. The pgaudit module now allows monitoring the activity of users not on the “whitelist,” and for trusted users, an option to disable logging in temporary schemas has been added. The transp_anon module implements anonymization at the table access method level, which prevents bypassing protection.

A specialized version for the “1C:Enterprise” platform has also been presented. Tests on an industrial 1C:ERP database of 646 GB showed a reduction in cost calculation time by more than 25% due to approximately 50 internal optimizations. These include a partial aggregation mechanism before JOIN execution and improved query selectivity estimation.