Though DAXA will at some point be able to stream data from the S3 bucket, there will still be a necessity to store some data locally (i.e. on a Fornax home directory), if raw XMM data are processed into cleaned event lists etc.
I've always erred on the side of keeping everything, but an effort should be made towards retaining only essential outputs to save storage (expensive on AWS).
First target is XMM processing - don't need to keep the non-final event lists, that could save a decent fraction of the storage consumed per ObsID (though observation length will determine the absolute size of that saving).
Though DAXA will at some point be able to stream data from the S3 bucket, there will still be a necessity to store some data locally (i.e. on a Fornax home directory), if raw XMM data are processed into cleaned event lists etc.
I've always erred on the side of keeping everything, but an effort should be made towards retaining only essential outputs to save storage (expensive on AWS).
First target is XMM processing - don't need to keep the non-final event lists, that could save a decent fraction of the storage consumed per ObsID (though observation length will determine the absolute size of that saving).