Trino performance on S3-compatible storage

[collimarco]

I am looking at Trino for large datasets stored on S3-compatible storage and I have some doubts:

1. Can it cache locally (e.g. on ssd or ram) the files read from S3? Or it reads them every time from remote storage?
2. Does it need to read the entire files or it can read only the relevant parts of the files? How?

Thanks