How do I find the maximum possible throughput for a single SilkPerformer Agent machine?
To identify maximum possible throughput (hits per second on server for example) do a stress test with say 10 - 20 concurrent users requesting a small single file of about 500Bytes to 1kByte. In this way, the agent will probably be the bottleneck. If you use bigger files (eg. 50k), the network bandwidth will often become the limiting factor.
In these circumstances the number of hits achieved per second on the server will give a good indication of the maximum throughput of the Performer Agent machine.
Note that parsing will affect this value and therefore you will get a higher number of hits with the browser level API than with the page level API. The difference will decrease with more cpu power.