Changes in HeatWave 9.3.1 (2025-05-27, General Availability)

Note

These release notes were created with the assistance of HeatWave GenAI.

• HeatWave GenAI

• HeatWave Lakehouse

• HeatWave MySQL

HeatWave GenAI

• HeatWave GenAI is now available on all HeatWave Cluster shapes, including the HeatWave.Free shape.

  Also, the following in-database LLMs are now available in HeatWave GenAI: llama3.2-3b-instruct-v1, llama3.2-1b-instruct-v1, llama3.1-8b-instruct-v1, and mistral-7b-instruct-v3. With the introduction of new LLMs, HeatWave GenAI now uses llama3.2-3b-instruct-v1 as the default LLM for content generation instead of mistral-7b-instruct-v1.

  For more information, see Supported LLMs, Embedding Models, and Languages. (Bug #37551135, WL #16509)

HeatWave Lakehouse

• HeatWave Lakehouse now supports seamless loading of highly-encoded Parquet files, including those with sparse data, highly repetitive values, and consecutive values. (WL #16812)

• HeatWave Lakehouse now supports unified loading of files using different URI formats. With this enhancement, you can load files from Object Storage using any of the three URI formats: OCIFS URI, PAR URI, and native URI. These URIs support patterns, prefixes, or names, which makes it easier to work with large datasets. Additionally, error and warning messages have been improved to include the relevant URI information for better troubleshooting.

  For more information, see Uniform Resource Identifiers. (WL #16822)

• HeatWave now supports exporting query results to Object Storage as Newline Delimited-JSON (ND-JSON) files. Using ND-JSON files lets you export query results as line-separated JSON logs, which enables easy and efficient integrations of your data with other web applications. This enhancement also makes it easier to load the exported results into a table with a single column of JSON data type, or use the HEATWAVE_LOAD routine to load the exported results into HeatWave Cluster for further processing.

  For more information, see Exporting Query Results to ND-JSON Files. (WL #16778)

• HeatWave now supports MEDIUMTEXT, LONGTEXT, and JSON data types with a column width of up to 4MB for Lakehouse tables. With this enhancement, you can leverage the expanded capabilities of HeatWave Lakehouse to load and manage larger datasets, boosting overall performance and productivity.

  For more information, see Lakehouse Limitations for all File Formats. (WL #16472)

HeatWave MySQL

• HeatWave now supports automated partition management, extending the Auto Parallel Load feature of automatically loading and unloading only the necessary partitions in partitioned tables based on usage. This enhancement accelerates workloads on partitioned tables while offering memory savings, allowing for more efficient data management and improved performance.

  For more information, see Automatic Loading and Unloading of DB System Tables and Paritions. (WL #15994)

• HeatWave now supports creating asynchronous tasks for long-running queries and commands, allowing them to run in the background as separate tasks. This enhancement enables you to run complex queries without blocking other operations, thus enhancing the overall performance and usability of HeatWave.

  For more information, see Running Tasks Asynchronously. (WL #16539)