CoS schedulers shaping and drop profiles

Packets are now classified and marked, but the WAN link is the bottleneck: Northmere pays for 10 Mbps of a 1 Gbps port, and the provider drops what's over. You'll shape R1 to the contract, so that R1 decides what's dropped rather than the provider, and read what happens in each queue under real overload.

What this lab can and can't show. vJunos has a software forwarding plane, so queueing isn't the same as on hardware, and we've measured how it differs. Shaping is real, and so are the per-queue counters and tail drops, but only with flexible-queuing-mode on. Under congestion, the lowest queue number goes first. transmit-rate, priority, exact, buffer-size and drop profiles all commit and show correctly, but in our tests they didn't change what was forwarded or dropped. You'll configure them as the exam expects, and the text says where a real MX or EX would behave differently. Don't assume these results carry over to hardware.

R1 starts with the classifier and rewrite rule from the last lab, with voice moved to queue 0. It also has chassis fpc 0 flexible-queuing-mode. Adding that statement to a running router restarts its forwarding plane, which took over 4 minutes in our tests, so it's in the start-up configuration here.