Database Technology

CORTO deployed Amazon Aurora PostgreSQL with pgvector to handle 7.6 billion vectors and 2.5 billion documents in production. The system delivers sub-second semantic search results for over 10,000 law firms while reducing storage costs by 75%. This demonstrates the capability of managed PostgreSQL for large-scale vector workloads.

• Aurora PostgreSQL with pgvector supports billions of vectors in production environments.
• Legal semantic search achieved sub-second latency for 10,000+ firm users.
• Storage costs reduced by 75% compared to previous implementations.
• Validates Aurora as a viable platform for massive vector indexing needs.

PostgreSQL 18.4 users report up to 23x faster multi-row insert performance by switching primary keys from UUID v1 to v7. The change requires an exclusive lock during the ALTER TABLE operation, which the author mitigated using short lock timeouts and retries. While UUIDs provide global uniqueness, this performance gain highlights v7's time-ordered structure as a superior alternative to v1 or v4 for high-throughput workloads.

• UUID v7’s time-ordered layout reduces index fragmentation and insert contention compared to v1/v4.
• Switching defaults requires an exclusive lock; use lock timeouts and retries to minimize downtime.
• Performance gains were measured on tables with billions of rows under heavy write load.
• PostgreSQL 18.4 is required to fully leverage UUID v7 capabilities for this optimization.

AWS has purchased DuckLabs, the company responsible for developing and supporting DuckDB, a popular in-process analytical database. This acquisition signals Amazon's commitment to the tool and its developer team as it integrates the technology into its cloud ecosystem. The move brings the core maintainers of the open-source OLAP engine directly under Amazon's corporate umbrella.

• DuckDB is now an AWS-owned asset, likely accelerating cloud-native integration features
• In-process architecture allows low-latency analytics without external server dependencies
• Expect deeper AWS service integrations for real-time data processing workloads
• Open-source license terms remain unchanged despite corporate ownership shift
• Developers should monitor for new managed hosting or optimization options from AWS

Christophe Pettus explains that PostgreSQL's logging collector functions as a small pipe-based daemon designed to prevent message loss and garbling. Enabling this feature requires a full server restart, a constraint practitioners must account for during configuration changes. The post clarifies the operational implications of using this mechanism for reliable log management.

• Logging collector uses a pipe-based daemon to ensure message integrity.
• Enabling logging_collector mandates a server restart.
• Restart requirement impacts change window planning for DBAs.
• Use this mechanism to prevent log message loss or garbling.
• Plan maintenance windows before toggling this specific GUC.

PostgreSQL 19 introduces the WAIT FOR command, enabling sessions to block until WAL reaches a specific Log Sequence Number. This mechanism provides read-your-writes consistency on asynchronous replicas, eliminating the need for the performance overhead of synchronous replication. The syntax supports configurable modes, timeouts, and error handling options like NO_THROW.

• Enables read-your-writes consistency without synchronous replication latency penalties.
• New SQL command blocks session until WAL reaches a specified LSN position.
• Supports MODE, TIMEOUT, and NO_THROW options for flexible error handling.
• Reduces operational complexity for applications requiring strong consistency on replicas.
• Requires upgrade to PostgreSQL 19 to utilize the WAIT FOR functionality.