I'm curious if anyone has experience auto-scaling a multi-tier ArcGIS Enterprise on Azure or AWS? If so, what instance types do you use for your Federated GIS Server?
I see in the Enterprise Cloud Builder for Azure that Esri defaults to D2_v2's (2 core 7GB RAM) for the 2 Web Adapters, and then DS3_v2's (4 core 14GB RAM) for the Portal, ArcGIS Server, and Relational Data Store.
I had initially been looking at the latest computing instances, namely AWS c5.xlarge (4 core and 8GB RAM) or Azure's F4s_v2 that advertise really fast data retrieval speeds. These have less RAM per core than the general compute instances, but they are much cheaper, and I figured with auto-scaling the RAM bottleneck could be mitigated.
Thanks for any help you can provide,