Changes in HeatWave 9.0.0 (2024-07-02, General Availability)

• HeatWave AutoML

• HeatWave Autopilot

• HeatWave GenAI

• HeatWave Lakehouse

• HeatWave MySQL

HeatWave AutoML

• HeatWave AutoML has increased the model size from 900MB to 4GB. (WL #15931)

HeatWave Autopilot

• HeatWave Autopilot includes improvements to cardinality estimation for queries that run in HeatWave MySQL and HeatWave Lakehouse. This improves the query plan, and provides improved performance (WL #15995)

• HeatWave Autopilot now includes Auto Indexing Advisor that offers indexing suggestions to improve the performance and latency of OLTP workloads running on the HeatWave MySQL primary engine. (WL #15663)

HeatWave GenAI

• HeatWave GenAI introduces similarity search capabilities that support the DISTANCE function that measures the distance between two vector values. It can include an optional distance_metric that can have a value of DOT, COSINE or EUCLIDIAN. This is available in the primary and secondary engines. (WL #16081)

• HeatWave GenAI introduces the new native VECTOR data type and in-database vector store support. This provides a fast, end-to-end, fully integrated pipeline which automates the vector store creation:

  • Reading unstructured data in PDF, HTML, TXT, PPT or DOCX format from object store.

  • Parsing the text in the documents.

  • Partitioning the text into smaller paragraphs and segments.

  • Encoding the paragraphs.

  • Storing the encoded paragraphs in a standard MySQL table in HeatWave.

  (WL #16106)

• HeatWave GenAI introduces a set of intuitive chat functionalities that enable enterprises to quickly build their own AI powered chatbot. It also introduces an out-of-the-box chatbot – HeatWave Chat, which integrates the chat functionalities that enable users to interact with HeatWave using natural language in addition to SQL. (WL #16248)

• HeatWave GenAI introduces in-database LLM with support for the following LLMs, see HeatWave GenAI:

  • Mistral-7B-Instruct.

  • Llama-2-8B-Instruct.

  • LLMs from OCI GenAI.

  (WL #15844, WL #16145)

HeatWave Lakehouse

• HeatWave Lakehouse now supports unstructured documents, and loads them to the GenAI vector store. Supported formats include PDF, HTML, TXT, PPT and DOCX. (WL #16100)

• HeatWave Lakehouse introduces Lakehouse Incremental Load that uses change propagation to refresh data views following changes to a loaded table. (WL #15817)

HeatWave MySQL

• HeatWave MySQL includes improvements to query optimization that lead to better performance for low latency queries in HeatWave MySQL and HeatWave Lakehouse. (WL #16222)

• HeatWave MySQL includes improvements to change propagation performance. (WL #16148)

• HeatWave MySQL improves the debugging of offloaded queries. (WL #15988)

• HeatWave MySQL now supports the use of cursors from stored procedures. (WL #16142)

• HeatWave MySQL now has support for the JSON data type with the following functions:

  • Cast Functions and Operators.

  • COALESCE() and IN(), see: Comparison Functions and Operators.

  • Control Flow Functions and Operators.

  • Mathematical Functions.

  • String Functions and Operators.

  (WL #16048)

• HeatWave MySQL now supports the following JSON functions:

  • JSON_ARRAYAGG()

  • JSON_CONTAINS()

  • JSON_CONTAINS_PATH()

  • JSON_KEYS()

  • JSON_OBJECTAGG()

  • JSON_OVERLAPS()

  • JSON_PRETTY()

  • JSON_QUOTE()

  • JSON_SEARCH()

  • JSON_STORAGE_FREE()

  • JSON_STORAGE_SIZE()

  • JSON_TYPE()

  • JSON_VALID()

  • JSON_VALUE()

  • MEMBER OF()

  (WL #16046, WL #16047)

• HeatWave MySQL now includes support for GROUP_CONCAT() with DISTINCT optimization and the ORDER BY clause. (WL #15906)

• HeatWave now supports dynamic query offload which uses machine learning based on query properties to determine the best engine, InnoDB or HeatWave analytic engine, to process the query. (WL #15829)

• HeatWave MySQL now supports concurrent query execution that reduces query latencies, and increases resource utilization. (WL #14826)