The tiered storage project has assumed that we want to store relatively large objects (10s of MB) in the cloud.  The goal of this ticket is to unpack that assumption by looking at how object size affects the cost and performance of using S3 and seeing how this might inform our design.