Discussion:
MIT's bufferbloat class
(too old to reply)
Dave Taht
2018-09-20 22:24:39 UTC
Permalink
It's a bit flawed - not using https, nor a reasonable number of http
assets, averaging the http result rather than plotting it also, with
min/max... nor using flent, nor publishing their source widely so that
I could fiddle with setting up mininet to verify it produces correct
results...

but, it's the first quiz on a class at MIT and it's hard to complain
more than I just did. :) I'm very tempted to try to do this one on my
own, as I strongly suspect the http result to be very misleading.

http://web.mit.edu/6.829/www/currentsemester/psets/pset1/ps1.html

* Start a long lived TCP flow sending data from h1 to h2. Use iperf.
* Start back-to-back ping train from h1 to h2 10 times a second and
record the RTTs.
* Plot the time series of the following:

*The long lived TCP flow's cwnd
* The RTT reported by ping
* Queue size at the bottleneck

Spawn a webserver on h2. Periodically download the index.html web page
using h1 and measure how long it takes to fetch it (on average).
Repeat the above experiment and replot all three graphs after enabling
the PIE AQM algorithm on the switch interface attached to the
bottleneck link. Set the target delay for PIE to 20 ms.
Repeat the above experiment with PIE enabled, but this time, start 10
parallel TCP flows with iperf.
--
Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619
Loading...