Thanks @tp_atkins we actually caught up with SAFE at the Esri UC and determined that while technical workarounds such as the one you suggested exist on the web and may functionally work, as per licensing agreements (Esri & SAFE), the requirement is for ArcGIS Server and all other setups are non-compliant. Bit rough because it instantly puts a big multiplier on FME server total cost of ownership (TCO). This is driven by the fact Esri licenses Server for ArcGIS on a 4-core minimum basis per machine so that’s a big cost to swallow just to get access to the ArcObjects DLLs, ArcPy Module and licensing mechanisms of Esri.
SAFE also confirmed that the ArcServer.exe doesn’t need to be running. While that at least means the server doesn’t need to be scaled in terms of RAM/CPU resources to have the software sitting idle it does raise bigger picture enterprise architecture questions that will be more easily answered by large organizations but becomes a challenge for smaller organizations. In a larger organization landscape, there may already be an ArcGIS Server established for data processing needs (ETLs, batch processes, etc.) where significant system resources have already been built and the ArcGIS Server licensing scaled for ‘x’ cores, and it makes sense to simply add FME Flow (Server) to that machine to maximize resource utilization. In a smaller organization, there likely isn’t a dedicated data processing ArcGIS Server and those needs are handled by the same stack that is handling general services needs for web maps & apps. FME is usually on its own machine and may have scaled resources to meet the load of data processing needs. This leaves smaller environments with a cost/benefit analysis to solve but neither equation is overly palatable:
- Use existing ArcServer for visualizations, add FME Flow to that but be forced to scale resources and hope that your needs are for just 1-2 additional CPU cores because at 3+ additional cores your ArcServer per core licensing costs fall towards equation 2. This way you benefit/maximize your existing ArcServer license costs and try to minimize any additional licensing impacts.
- Hope your existing stand-alone FME Flow server has no more than 4 cores, install ArcServer to it and set the ArcServer.exe to be a ‘Stopped’ service. You still eat the full cost of the ArcServer license and get zero use/benefit from the license outside of being able to use FME with Esri solutions, but you don’t overtax your visualization server and perhaps can do some rebalancing across your architecture (could be networking, security group, authorization, etc. impacts to a change and are those headaches/costs worth it) to shift any processes that had been running on the visualization server to the FME server, making it the start point for data processing. Obviously, a lot easier decision if you are already doing a full environment upgrade/migration AND your FME footprint is within the purview of that larger consideration of an Esri/GIS stack vs. trying to sort it all out on existing infrastructure and at best make a band aide solution with a bookmark for it in some future effort.
All things considered, we must recommend the compliant path, and our client only wanted to consider compliant options. It is without doubt a rock and hard place scenario with a zero-sum answer...your licensing costs will go up. Since the FME server environments (DEV, TEST, PROD) were well established and not in scope, and each are beefy systems for obvious data processing reasons, the decision was to install ArcGIS Server to these servers under the existing EA and recognize its at best an item to resolve via re-architecture & optimization before the next Esri EA renewal and at worst a tough conversation point during the renewal.
