I'm confused about how Snowflake handles storage billing for Zero Copy Cloning. If I clone my production database to a dev environment and then run massive update scripts in dev, am I only charged for the modified data or does it create a full billable copy of the database?
3 answers
Zero Copy Cloning is one of Snowflake's best features because it doesn't actually replicate data at the physical storage layer initially. When you clone, the metadata just points to the existing micro-partitions. You are only charged for "new" storage when you modify data in the clone. For example, if you update 1GB of data in a 1TB cloned database, you only pay for that 1GB of unique data plus the original 1TB from the source. It’s a huge win for DevOps because you can test production-scale data without the production-scale storage bill.
Are you planning to keep these clones long-term, or are they ephemeral environments that you will drop immediately after your deployment tests are finished?
Just remember that while storage is nearly free with clones, compute is not. Running a query on a clone costs exactly the same as running it on the original source table.
Good point, Susan. A lot of people forget that they still need a warehouse to actually look at the cloned data, which is where the real Snowflake costs usually live.
Thomas, we plan to refresh them weekly. My main worry is the 'Time Travel' retention settings on the clones. If we drop the clone, does the modified data go into Fail-safe storage, and will we be billed for that storage for 7 days? We move about 500GB of data per day, so those leftovers could add up quickly if we aren't careful with our cleanup scripts.