Save yourself the trouble: the box will do line-rate on all ports, down to as small packets as you want to push; all in a single flow, or stepping through single-packet flows. If I recall, there are a few particular packet size thresholds that could affect performance, and TCAM space is not infinite, but there is nothing that will affect your real-world forwarding.
Exception Path Processing and the umbilical ethernet to the routing engine were the only interesting limitations I recall on the previous-generation Junipers. There's that whole "out-of-order packet" issue with the M160, but I don't thing that in-order delivery is an issue beyond the datacenter/enterprise; not for backbone routing.
Speaking of "switch", yes, the M160 has been used as a switch: a giant frame-relay switch, for a carrier which couldn't get beyond OC48 rates on the Marconi ASX4000 platform. I think PacBell NAP also tried using it to extend their ATM peering point to 10G (lotta good it did them).
Forwarding is not all a router does, and Juniper isn't perfect. But you won't find anything interesting with your performance test if you stick to core-router applications.
|