Hello!
This is a very pertinent question, however as many things in life there is no one size fits all here.
We basically recommend, as best practice, to allocate for each specific context only the estimated needed resources. These values should always come from a previous study on the network patterns/load.
To accomodate for growth and scalability it is strongly advised to initially keep as many resources reserved as possible and allocate the unused resources as needed. To accomplish this goal, you should created a reserved resource class, as you did already, with a guarantee of 20 to 40 percent of all ACE resources and configure a virtual context solely with the purpose of ensuring that these resources are reserved.
As you might already know ACE protects resources in use, this means that when decreasing a context's resources, the resources must be unused before then can be reused by other context. Although it is possible to decrease the resource allocations in real time, it typically requires additional overhead to clear any used resources before reducing them.
Based on the traffic patterns, number of connections, throughput, concurrent SSL connections , etc, for each of the sites you will be deploying you will have a better idea on what might be the estimated needed resources and then assign them to each of the contexts. Thus this is something that greatly depends on customer's network environment.
Hope this helps to clarify your doubts.
