Bufferbloat

``Bufferbloat'' is the growth in buffer size that has led Internet delays to occasionally exceed the light propagation delay from the Earth to the Moon. Manufacturers have built in large buffers to prevent losses on Wi-Fi, cable and ADSL links. But the combination of some links' limited bandwidth with TCP's tendency to saturate that bandwidth results in excessive queuing delays. In response, new congestion control protocols such as BitTorrent's uTP/LEDBAT aim at explicitly limiting the delay that they add at the bottleneck link. We propose, validate a methodology for bufferbloat inference, and present results from an Internet campaign, making our results, code and datasets available to the scientific community.

Experimental validation

We propose a methodology to monitor the upstream queuing delay experienced by remote hosts, both those using LEDBAT, through LEDBAT's native one-way delay measurements, and those using TCP, through the Time-stamp Option. The methodology has been published at PAM'13, demonstrated at P2P'XII and implemented in Tstat (the code is available upon request).

We extensively validate the methodology in a testbed. We report the raw packet level traces and kernel and application level ground truth logs that we used to validate the methodology. Traces are exactly those used in this extended technical report to obtain Fig. 2(a) and (b) shown above, and file names in the dataset are labeled accordingly.

Internet experiments

In the attempt of assessing the bufferbloat in the wild, we have performed an Internet measurement campaign, whose data is available below. Our measurement campaign let us conclude that (i) LEDBAT delay-based congestion control is effective in keeping the queuing delay low for the bulk of the peers, (ii) yet about 1\% of peers often experience queuing delay in excess of 1 s, and (iii) not only the network access type, but also the BitTorrent client and the operating system concurr in determining the bufferbloat magnitudo.

We make the dataset used in TMA'13 paper available to the community. Dataset consists of several BitTorrent packet-level traces, concerning 12 torrents, captured from 3 vantage points for a total of 88 experiments in which we contacted over 25K external peers.

  • Internet experiment (20GB compressed size, 88 experiments, raw pcap traces captured)
  • Four peers sample used in Fig2 of TMA'13

References

  1. Chiara Chirichella, Dario Rossi, To the moon and back: are Internet bufferbloat delays really that large . In IEEE INFOCOM, Traffic Measurement and Analysis Workshop, April 2013.
  2. Chiara Chirichella, Dario Rossi, Claudio Testa, Timur Friedman,Antonio Pescape, Remotely Gauging Upstream Bufferbloat Delays . In Passive and Active Measurement (PAM), Poster session, extended technical report, march 2013.
  3. [P2P-12] Chiara Chirichella, Dario Rossi, Claudio Testa, Timur Friedman, Antonio Pescape, Inferring the buffering delay of remote BitTorrent peers under LEDBAT vs TCP . In IEEE P2P'XII, Demo Session, september 2012.
  4. Chiara Chirichella, A methodology to gauge the extent of bufferbloat in the Internet . In MSc Thesis, september 2012.