Due to the lack of portal view layers for registered feature services (or the ability to proxy access to restricted layers through a map's/app's sharing options) I often have to create multiple feature services with minor adjustments to ensure the correct group gets the correct access. This leads to the following dilemma:
- If every service is dedicated with 1 minimum instance, there's a high static load even if a small number of user types are accessing the data.
- Cut the minimum count to 0 and now every new group has a terrible experience as their service fires up. Setting the time this instance stays alive is an impossible tightrope act as user activity varies day by day, so it's either active too long or inactive too often.
- The shared pool is unsuitable for this type of service as this project is now at war with every other project.
My proposal: allow admins to create their own shared pools (with their own max core counts, max active services etc.) and allow users to choose if they want to publish to these shared pools or the default one. Admins can now balance server load at the pool level and the pool can balance service access using its work balancing methods. This isn't a trivial addition to the platform but I think the efficiency gains in larger deployments are well worth it.
