There are a lot of things that affect BigFix performance. Part of it is definitely the number of endpoints, but the number of console operators, particularly the number that will use it all at the same time at peak times has an impact.
If you have less than 10 simultaneous operators and less than 10000 endpoints and don’t expect a lot of growth, then it takes a lot less effort to have something that works very well. Whenever you hit numbers much higher, then some of the software and hardware configuration starts to matter a lot more.
I don’t think Amazon AWS or Azure are officially supported, but I do know some smaller infrastructures are using them. You may be taking a risk as far as support goes if you use something that isn’t officially supported.
There are a few different parts of BigFix that are sensitive to the storage IOPS and latency while most of it is much less so. IOPS matter for the Databases, FillDB, WebReports Cache, WebUI Cache, Console Caches, and anything similar. Meanwhile the download / mirror caches are more of bulk storage so they can be on larger and slower (by comparison) storage.
How many clients and operators do you expect to have?