Oracle is expanding its HeatWave cloud database service with a suite of new generative AI services called HeatWave GenAI. This platform, previously known as MySQL HeatWave, offers an extended, cloud-managed version of the MySQL database with transactional and analytical capabilities. Last year, Oracle introduced the HeatWave Lake House to provide data lake house functionalities.
HeatWave GenAI integrates vector processing and advanced AI functions into the database. Oracle is among the few vendors incorporating Large Language Models (LLM) directly into their database, offering quantized versions of Call 3 and the LLM from Mistral. This integration aims to enhance performance and introduce new applications that complement the existing AutoML capabilities of HeatWave.
According to Nipun Agarwal, senior vice president of MySQL and HeatWave at Oracle, having LLM directly within the database eliminates the need to provision a GPU externally, resulting in better synergy with AutoML and other HeatWave features running in the database.
The introduction of LLM in databases is a first in the industry and complements Oracle’s existing AutoML feature. For example, early users have utilized HeatWave GenAI for anomaly detection and making recommendations for online food delivery businesses. Agarwal believes this combination will generate new applications and value for customers.
Automated vector processing is another feature of HeatWave GenAI, which includes a built-in Vector store. While vector support in databases has become common, Oracle’s vector embedding feature automates the generation of vector representations for unstructured documents like text, images, and videos, eliminating the need for manual expertise in data conversion and processing optimization.
Unlike other database vendors that rely on approximate indexing for vector search, HeatWave GenAI emphasizes in-memory table-scan operations for precise performance. Agarwal highlighted that using indices for performance measurement may sacrifice precision, which Oracle aims to avoid with their approach.
In summary, Oracle’s HeatWave GenAI brings advanced AI capabilities, vector processing, and LLM integration to its cloud database service, offering unique features that enhance performance, enable new applications, and provide value to customers. The platform’s automation of vector processing and focus on precise in-memory operations distinguish it from other database offerings in the market.
Article Source
https://venturebeat.com/data-infrastructure/oracle-bakes-llms-and-vector-support-directly-into-heatwave-genai-database/