TEST OF CAR WITH SPECIFIC NORMAL AND EXCEED BURST CONFIGURATION

TEST A


"unconfigure CBWFQ on output and do not send any UDP traffic. Just measure the TCP goodput when TCP is alone and just the policer is running. I expect that the goodput will be well below 1.3Mbps due to the known mismatch between a strict token bucket and a bursty TCP flow I expect that the higher the number of flows (thus the more realistic the test), the better the goodput" (C.Filsfils) Router configuration:
NOTE: "router output" is such that before the start of each test the command "clear counter" has been issued to reset the packet counters.
TEST B
"reconfigure CBWFQ and resend UDP traffic but with: NB = 1296000 * x ; EB = 1296000 * 2 * x
where x = 50ms (your test), 250ms, 500ms, 1s, where NB = normal burst, EB = excess burst
I expect that the goodput will improve and approach 1.3M when x will increase to 1s" (C.Filsfils) Router output: