If you're a Catalog Curator or Admin, you know the tension: users want more data in the marketplace, but every new dataset means more storage, more sync jobs, more staleness risk, and more governance overhead to track. Zero-copy (virtualization) removes a big chunk of that burden.
What it solves for you
- No duplication to manage. Zero-copy assets keep their records in the source system (Snowflake or Databricks). You expose the data without ingesting and storing a copy in Huwise : one less thing to maintain, monitor, and pay for.
- Data is always current. Because exploration queries hit the source directly, there's no sync schedule to configure, no "last updated" gap to explain to users. What they see in the marketplace is what's in the warehouse, right now.
- Governance stays exactly where you expect it. Zero-copy assets go through the same publication workflow as any other harvested asset : nothing is published automatically, so you still review before making it live. Access control and security sanitization on real-time queries are maintained at the same level as standard datasets, so you're not trading governance for convenience.
- Faster time-to-catalog. Since Snowflake and Databricks harvesters can mass-import and virtualize assets in bulk, you can expose a large slice of a warehouse or lakehouse to your marketplace in one harvester run (instead of provisioning storage and running individual imports asset by asset).
Where the trade-off is
Zero-copy isn't a full replacement for copied datasets yet. Explore by Query, AI Insights, export, the API console and custom views aren't supported on virtualized assets today. As a curator, this is the main thing to factor into your decision:
- if end users need those capabilities on a given dataset, keep it as a standard copy;
- if they mainly need to search, filter, sort, and understand the schema, zero-copy is a lighter, faster path to publication.
One nuance for admins configuring sources
The Snowflake/Databricks harvesters virtualize assets in bulk, while the Snowflake/Databricks connectors (for individual imports) still copy data. So the choice of tool determines the behavior, not the source itself.
💡 Zero-copy is an add-on, reach out to your Account Manager or Customer Success Manager to activate it.
📖 Full documentation: Zero-copy data sharing (virtualization)