March 2026 Summaries

AI deployment in enterprise environments is often hindered by challenges related to data access and data governance. While technology plays a significant role in AI success, contextual understanding of data is crucial, with poor data access and lack of governance being major obstacles. Universal data access is posited as a solution, allowing organizations to choose how they access and govern data without defaulting to centralization. This approach balances local control with centralized monitoring and audit capabilities, facilitating AI readiness. Adopting a hybrid AI data strategy, especially in regulated industries, ensures compliance while leveraging both cloud and on-premises data. Data products are emphasized as key enablers, providing curated datasets that enhance data governance and support AI operations. These strategies, along with tools like the Starburst AI Data Assistant, aim to replace traditional BI dashboards with interactive AI interfaces, aligning AI initiatives with business goals and fostering stakeholder trust.

Starburst Enterprise has introduced the Great Lakes Connector, a unified connector designed to streamline the management of multiple data lakehouse table formats, such as Hive, Iceberg, Delta Lake, and Hudi, in enterprise environments. By consolidating these formats into a single connector and catalog, it simplifies the often complex and lengthy migration processes that organizations face when transitioning between different data formats. This connector enhances operational efficiency by allowing seamless SQL queries across various table types without necessitating changes to existing object storage and metastore configurations. Built upon the high-performance architecture of Starburst Galaxy, it offers advanced features such as location-aware access control and accelerated data access through parallel streaming of metadata, positioning itself as a robust alternative to the open-source Trino Lakehouse connector. The Great Lakes Connector is particularly tailored for enterprise readiness, providing a consistent and simplified user experience that reduces the total cost of ownership and focuses on delivering data value rather than managing infrastructure complexities.

Starburst's Rewind and Backfill feature offers data engineers a robust solution to address common challenges in streaming data pipelines, such as schema drift and parsing logic issues. Built on the Iceberg platform, this tool enables engineers to "rewind" an Iceberg table to a previous state before errors occurred and then "backfill" the data using updated logic, ensuring the integrity and consistency of the data without creating duplicates or losing information. The process is straightforward and involves updating parsing logic, selecting a prior point in time, and triggering a backfill operation, which recomputes data accurately. This capability allows for seamless iterations and corrections, eliminating the need for complex manual backfills. The underlying architecture, which includes exactly-once processing, a managed control plane, and incremental materialized views, ensures efficient and reliable data management. By using a scalable and elastic multitenant architecture, Starburst can handle high-speed data ingest, making Rewind and Backfill a powerful tool for maintaining data quality in dynamic environments.

The Starburst Enterprise Intelligence Platform, running on the NVIDIA Vera, was benchmarked against Intel Xeon 6 and AMD EPYC Zen 5 platforms to evaluate modern server CPU performance for large-scale SQL analytics workloads. Utilizing the TPC-DS benchmark and Apache Iceberg tables, the study highlighted that NVIDIA Vera provided approximately three times faster query throughput and up to 1.85 times better CPU efficiency than Intel Xeon, while maintaining comparable efficiency with AMD Zen 5. The benchmark involved executing a variety of analytical SQL queries that emphasized scan, join, and aggregation operations, revealing significant performance advantages of NVIDIA Vera in sustaining parallel execution and improving infrastructure efficiency. These improvements are critical as datasets scale into terabytes and petabytes, underscoring the importance of CPU architecture in modern analytics and AI workloads.

Starburst has announced that it is the first AI platform to support NVIDIA Vera, marking a significant step in AI-driven enterprise intelligence. The Starburst Model Context Protocol (MCP) Server facilitates a shift from traditional business intelligence (BI) to AI-powered insights by providing a secure, standardized interface for AI agents to access governed and federated data in real-time. This platform alignment with NVIDIA Vera enables rapid AI-based analytics, leveraging Vera's optimized inference engine for efficient processing. The integration aims to transform how data is consumed by moving beyond static dashboards to a dynamic, conversational model where AI agents can access structured business contexts, ensuring accuracy and governance at scale. The collaboration addresses performance bottlenecks and governance challenges, enabling seamless, federated data access and deterministic throughput for AI tasks. With its enterprise-grade governance and performance capabilities, the Starburst MCP Server and NVIDIA Vera platform are designed to support large-scale AI adoption, turning complex data estates into accessible, governed resources that align AI ambition with operational realities.

Starburst Galaxy's latest release marks a significant shift from traditional business intelligence to AI-driven analytics by leveraging the Icehouse architecture and introducing the AI Data Assistant (AIDA). This development promises to enhance data interaction through conversational, agentic interfaces, moving away from static dashboards and enabling continuous, dynamic analysis. Built on a foundation of Apache Iceberg and Trino, the platform emphasizes rapid and efficient data ingestion, offering capabilities like high-velocity streaming and automated file ingestion from Amazon S3. AIDA operates on Galaxy's curated, governed data products, ensuring consistent metrics and robust data governance. The platform's performance is validated by independent benchmarks, showing superior ingestion rates and cost efficiencies compared to competitors. With new features such as Avro support for streaming ingest and expanded schema management, Starburst Galaxy is poised to facilitate a seamless transition to AI-powered decision-making, freeing users from the constraints of traditional data management and business intelligence tools.

Starburst Galaxy has demonstrated superior performance in data ingestion, significantly outpacing competitors like AWS Data Firehose and Confluent Tableflow, particularly in streaming data into Iceberg tables. This innovation addresses the common bottleneck in AI workloads by providing high-speed and cost-effective ingestion solutions, with features such as streaming ingest from Kafka sources and automated file loading from Amazon S3. Starburst achieves a 7x higher ingestion rate with an 87% cost reduction compared to AWS Data Firehose, confirmed by Concurrency Labs' benchmarking. The platform also excels in data compression, resulting in smaller Iceberg tables, which enhances cost efficiency and performance. Starburst is extending these capabilities to its Enterprise version by the end of the year, ensuring a unified, accessible, and manageable data lifecycle experience across its ecosystem.

Starburst Enterprise 479-e.1 LTS release enhances AI support and data management capabilities with significant improvements in SQL-based orchestration, engine performance, and Iceberg table maintenance. The release introduces enhancements like Common Table Expression (CTE) reuse for AI workloads, Multi-Statement SQL Jobs for streamlined query execution, and expanded connectivity through the Great Lakes and OpenAPI connectors. It also offers new features for sharing data products across clusters and improved infrastructure with Java 25.0.2 support. Key updates include mandatory security masking for custom connectors and changes to Apache Ranger support, as well as the removal of certain features in Starburst Warp Speed. These advancements aim to optimize resource utilization, ensure data accessibility, and maintain a secure and efficient data environment.