I've been looking into the most efficient way to carve up the allocations (in the resource-class's) for each context in a multi-tenanted environment.
I've been looking at deploying a resource-class per context. Each resource-class would allocate all resources at 20 to 30%. Within each context under the rserver I would rate-limit the bandwidth and connections to meet the requirements. And all other resources would be monitored. By having a resource-class per context allows you to increase or decrease resources should this be required depending on load.
What are your thoughts? Or would you create a context and allocate each resource a percentage based around the design requirements? For example, context B, requires 10Mb through put, 20,000 concurrent connections etc...
The guide should give you an idea of how to manage resources on ACE. Like you said you can always increase or decrease resources depending on the load.