Hello! I have a small question about managing s3 buckets. We are using intelligent tiering in AWS and it works great, saving us money, but are also thinking about using metaflow on GCP. Looks like GCP supports a similar feature called Autoclass Are there any other best practices for managing artifacts and their costs? I've seen some mention of this in past threads, so just curious if enabling Autoclass would be the state-of-the-art or if you have some other plans

hi! yes

autoclass

should just work (similar to intelligent tiering)

I would imagine that the biggest cost driver are large datasets stored as artifacts -- if they have not been accessed for X weeks or months then it would be safe to completely remove them (since the complete data still lives in our data warehouse)

one pattern would be to use metaflow.s3 to store data in a different bucket (following a different lifecycle policy) and storing the address as a metaflow artifact through

self.