although I've seen suggestions against modifying the default values for this command, I've come across a situation in which it seems to help a lot to increase the parameter "interrupt-time" associated with fast switching.
Basically, although the overall CPU util (on a 7200 w NPE-G1), is not exceeding 8% under this particular traffic profile the router drops packets when a "show run", "show ver" or "wr mem" commands are issued. The profile is roughly 100Mbps of unidirectional traffic, from which 30Mbps are packets of 128 bytes in size.
If I increase the mentioned parameter, there are no droped packets. However I'm conscious of the implications this could have with some other processes (i.e. routing protocols)
Few related questions:
How can we determine/calculate the right tunning of both values associated with the scheduler allocate command?
How can we relate this to the documentation which mentions "network interrupt contexts"? (Does this mean a single interrupt?)
How can this optimisation be related to traffic profiles? (i.e. interrupt time vs. packet rates, sizes, buffers, etc)