The Battle for Bandwidth: Fairness and Heterogenous Congestion Control on Today's Internet

Add to your list(s) Download to your calendar using vCal

• Justine Sherry, CMU
• Thursday 29 October 2020, 15:00-16:00
• https://meet.google.com/ehj-dwaz-rea.

If you have a question about this talk, please contact Srinivasan Keshav.

In recent years, the Internet has seen an explosion of innovation in the congestion control space. Where the Internet was previously dominated by a few well-studied algorithms (namely Reno-derivants and Cubic), today’s services may use one of a wide range of algorithms such as Copa, BBR , LEDBAT, GCC , etc. We often evaluate new algorithms from the perspective of the services that use the new CCA , e.g., looking at performance outcomes in terms of latency, flow completion times, or video rebuffering rates. However, less attention has focused on how these algorithms compete with each other when sharing a bottleneck link and whether or not new algorithms cause performance degradation for legacy services running side-by-side along the new algorithm.

In this talk, I will present a case study of a widely deployed algorithm—BBRv1—and how it behaves under competition with legacy algorithms. Our study of BBRv1 combines measurements and mathematical models to demonstrate that BBRv1’s bandwidth consumption is independent of the number of competing Reno or Cubic flows and is therefore fundamentally unfair to these existing algorithms; our results informed Google in the redesign of BBRv1, resulting in BBRv2. I will then explore a new methodology for deciding whether or not a new algorithm is acceptable for deployment on the Internet. Where the traditional literature has focused on fairness, we re-frame the discussion in terms of a concept called harm that we argue is more practical to serve as a standard or “bar” for what is acceptable to deploy. Our work on harm was awarded the IRTF Applied Networking Research Prize in 2019. Finally, I will present our new testbed for measuring the harm that Internet services inflict upon each other and some preliminary results from our testbed evaluating the harm caused by video services such as Netflix, Vimeo, and YouTube.

Bio: Justine Sherry is an assistant professor at Carnegie Mellon University. Her interests are in software and hardware networked systems; her work includes middleboxes, FPGA packet processing, measurement, cloud computing, and congestion control. Dr. Sherry received her PhD (2016) and MS (2012) from UC Berkeley, and her BS and BA (2010) from the University of Washington. Her research has been awarded the Applied Networking Research Prize, the SIGCOMM doctoral dissertation award, the David J. Sakrison prize, paper awards at USENIX NSDI and ACM SIGCOMM , and an NSF Graduate Research Fellowship. She is a member of the DARPA ISAT Study Group and the SIGCOMM CARES Committee. Most importantly, she is always on the lookout for a great cappuccino.

This talk is part of the Computer Laboratory Systems Research Group Seminar series.

This talk is included in these lists:

• All Talks (aka the CURE list)
• CL's SRG seminar
• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge talks
• Chris Davis' list
• Computer Laboratory Systems Research Group Seminar
• Department of Computer Science and Technology talks and seminars
• Interested Talks
• School of Technology
• Trust & Technology Initiative - interesting events
• bld31
• https://meet.google.com/ehj-dwaz-rea
• ndk22's list
• ob366-ai4er
• rp587

Note that ex-directory lists are not shown.