Session

Enterprise Cost Management for Data Warehousing with Databricks Lakehouse

Overview

ExperienceIn Person
TrackData Warehousing
IndustryEnterprise Technology
TechnologiesDatabricks SQL
Skill LevelBeginner
DOWNLOAD SESSION SLIDES

This session shows you how to gain visibility into your Databricks Lakehouse spend and ensure cost efficiency. Learn about the latest features for AI-driven insights into Lakehouse expenses so you can easily monitor and control your costs. Find out how to enable cost attribution to internal projects, improve performance and cost with agentic workload optimizations, and take advantage of intelligent compute and storage capabilities that do the tuning for you.

Session Speakers

Jiabin Hu

/Staff Software Engineer
Databricks

Joo Ho Yeo

/Product Manager
Databricks

Full Summary

A new playbook for governing Databricks SQL in the age of AI agents

Enterprise SQL usage has shifted from a few human analysts to a mix of BI tools and AI agents generating queries at machine speed. The result is more warehouses, more unoptimized SQL, and higher pressure to connect spend to business value. The session outlines a practical governance playbook that restores control without slowing teams down.

FAQ


No. Turn on the table-stakes features first, managed tables with predictive optimization and serverless SQL, because they improve both cost and performance with minimal tuning. Then decide whether to prioritize tighter cost controls or faster per-query performance.

Query Tags attach custom key-value metadata, such as cost center, project, dbt model, or Power BI report, to each SQL execution from any interface. The attributed usage system table exposes those tags, along with per-query DBU attribution, so you can aggregate and analyze spend by any tagged dimension.

It keeps small exploratory queries from waking large, expensive warehouses intended for ETL or heavy jobs. Admins set a designated default warehouse for ad hoc authoring so big clusters stay asleep unless a workload truly needs them.

Choose Lakehouse Real-Time, in beta, when you need sub-100-millisecond latency with very high concurrency directly on the governed lakehouse. If the requirement is mainly faster heavy queries, first scale t-shirt size or increase max cluster count on an existing warehouse.

Query Tags are in public preview and platform alerts are generally available. Query Performance Insights, Genie Code query rewrites, the attributed usage system table, and parts of the Performance tab in the Governance Hub are in beta or private preview, and the team encouraged feedback through account channels.