Zero-copy support for all table formats in interactive warehouses

Support for all table formats in interactive warehouses is now generally available. You can now execute low-latency and high-concurrency queries directly against your existing data assets without needing to copy or transform them into specialized interactive tables.

Use cases

Imagine you need to power real-time dashboards or high-throughput APIs leveraging your existing standard or Iceberg tables.

• Query standard, dynamic, Iceberg, and hybrid tables directly.
• Benefit from the low-latency and high-concurrency engine of interactive warehouses.
• Avoid duplicating or transforming data into separate tables.

How to use

You can create an interactive warehouse using specific keywords and associate tables for proactive cache warming.

CREATE OR REPLACE INTERACTIVE WAREHOUSE interactive_demo
  WAREHOUSE_SIZE = 'XSMALL';

What happens in this code:

1. Initiates the creation of a new compute resource.
2. Uses the INTERACTIVE keyword to target the specialized execution engine.
3. Sets the initial size of the warehouse to XSMALL.

To associate tables and trigger cache warming:

ALTER WAREHOUSE interactive_demo ADD TABLES (orders);

What happens in this code:

1. Alters the existing interactive warehouse.
2. Adds the orders table to the proactive cache-warming routine.

Limitations and requirements

• This feature is generally available (GA).
• Queries on interactive warehouses have a strict 5-second runtime limit; configuring a fallback warehouse is required for longer operations.
• The minimum auto-suspend interval is 24 hours (86400 seconds) to maintain cache consistency.
• You can attach up to 100 tables per interactive warehouse for proactive cache warming.
• CALL commands for stored procedures and the flow operator ->> are not supported.

Using it in a pipeline

This fits into the serving layer, allowing analytical apps and APIs to consume data from your data lakehouse tables with low latency and zero copy overhead.