Description
Vector databases such as ElasticSearch and Pinecone offer fast ingestion and querying on vector embeddings with ANNs. However, they typically do not decouple compute and storage, making them hard to integrate in production data stacks. Because data storage in these databases is expensive and not easily accessible, data teams typically maintain ETL pipelines to offload historical embedding data to blob stores. When that data needs to be queried, they get loaded back into the vector database in another ETL process. This is reminiscent of loading data from OLTP database to cloud storage, then loading said data into an OLAP warehouse for offline analytics. Recently, “lakehouse” offerings allow direct OLAP querying on cloud storage, removing the need for the second ETL step. The same could be done for embedding data. While embedding storage in blob stores cannot satisfy the high TPS requirements in online settings, we argue it’s sufficient for offline analytics use cases like slicing and dicing data based on embedding clusters. Instead of loading the embedding data back into the vector database for offline analytics, we propose direct processing on embeddings stored in Parquet files in…
Description from YouTube. Full content on the video page.
More from Databricks
NewsGovern MCP servers in Databricks #databricks #mcp #aigovernance
Databricks Unity AI Gateway now governs MCP servers, centralizing their management alongside built-in foundation models and LLMs. This integration allows for easier governance and orchestration of various AI components and agents within Databricks.
NewsHow Suntory Turns Data into Faster Decisions with Databricks
Suntory uses Databricks to integrate diverse datasets, including internal sales, macroeconomic factors, and consumer behavior, into "Project Brain" for faster decision-making and product launches. The company also implements an all-employee upskilling program, "Manabi no Michi," to empower its workforce to leverage AI for improved performance and efficiency.
NewsAIA Group x Databricks: Turning Regulated Data into Real-Time Intelligence
AIA Group leverages Databricks to manage regulated data across 18 markets, addressing challenges like data residency and varying tech maturity with features like Unity Catalog for governance. The platform enables real-time intelligence for investment decisions, fraud detection, and personalized agent coaching, with future plans for conversational analytics and autonomous AI.
TutorialsConnect Google Sheets to Databricks
The Databricks Google Sheets add-in allows users to explore, import, and refresh governed data from the Databricks Lakehouse directly within Google Sheets. It demonstrates how to browse Unity Catalog, select tables or metric views, apply filters, schedule data refreshes, and use direct SQL queries with parameters.
NewsNo More Table Locks for Multi Statement Transactions #databricks #dataengineering #sql
Databricks now supports multi-table transactions, allowing changes to multiple tables within a single atomic transaction that rolls back all changes if any part fails. This feature, managed by Unity Catalog, prevents table locking during updates and supports up to 100 tables per transaction using a simple "BEGIN ATOMIC...END" syntax.
NewsMay 2026 Databricks Updates: No Code ETL, New GPUs and Death of the Dashboard
Databricks announced several updates including AI Prep Search for document chunking and vector database preparation, SQL vector functions for embedding mathematics, and the general availability of multi-table transactions. They also introduced Lakeflow Designer for visual, no-code data pipeline creation and updated their serverless GPU offerings to include H100s.