Given that the requirements say that this service and the data it keeps needs to last through the current XSEDE period and most likely on to successors, are we certain that using Jetstream is a good idea? Have you confirmed that Jetstream's service lifetime extends as far as you need? My guess is it has about 3 more years, plus or minus 9 months? In any case, we know for certain that there will be an inevitable shutdown of Jetstream requiring us or our successors to migrate this service from there at some point.
Further, Jetstream is an NSF resource intended primarily for allocated scientific research (allocated by XRAS). At some point, some review panel (either Jetstream's or XSEDE's) is going to ask why Jetstream is being used to host other systems' critical services (as opposed to some other cloud provider whose primary business is to host things). It's probably "legal" (part of the Jetstream team's discretionary allocation), but the optics aren't very good IMO.
Speaking of other cloud providers whose primary business is to host things... Why wouldn't we host this on AWS, given that we're moving other things in that direction already? Is there any uncertainty about the availability of AWS for XSEDE (or its successors) in the foreseeable future?
Good point. We are in the process of figuring out how to manage our use of AWS: establishing base images to build from, automating image customizations using Ansible, establishing an XCI repo for Ansible recipes, documenting AWS operational processes, etc. We plan to establish these processes using first an Information Services instance, and then the CSR and other services. We'll include this service in our migration plan. Maybe this service could be consolidated into the CSR as a usage metrics element of the software repository?
Thanks.