RISE Seminar: Meeting stringent Internet performance requirements despite uncertainty, a talk by Sanjay Rao
May 10, 2019
Title: Meeting stringent Internet performance requirements despite uncertainty
Speaker: Sanjay Rao
Affiliation: Purdue University
Date and location: Friday, May 10, 12:30 – 1:30pm; Wozniak Lounge (430 Soda Hall)
Abstract: The Internet is the backbone of critical global cyber infrastructure,
and its continual phenomenal growth comes with increasingly stringent
expectations on performance. For instance, a recent paper from Google
indicates that traffic on their wide-area network infrastructure has grown
by 100X in the last 5 years, while performance requirements must now be
met 99.99% rather than 99% of the time. High application quality
(e.g., low latency Web applications, high video quality)
is critical to user engagement, and new demanding applications such as 4K and
360-degree video continue to emerge. The performance requirements must be met
despite significant uncertainty. First, the global scale of cloud provider
networks and the rapid evolution of these networks imply that failures are
the norm in production networks today. Second, from an application perspective,
users are behind diverse networks (e.g., wired vs. LTE vs. WiFi), whose
performance is often variable.
In this talk, I will present our recent research in tackling these challenges.
The first part is inspired by the vision of automating network design, an
ad-hoc process today that lacks formal guarantees. I will present a novel
optimization-theoretic framework that we have developed which allows
architects to formally reason about the performance of a network design over
all, or a percentage of, a large set of scenarios in which it must
operate. Our focus on quantitative performance properties distinguishes
our work from current efforts at network verification that have primarily
focused on correctness (e.g., ensuring reachability policies are met).
The second part will describe Oboe, a new system that we have
developed to auto-tune Available Bit Rate (ABR) algorithms for Internet video
streaming. Oboe achieves upto a 38% improvement in video delivery metrics over
the state-of-the-art including Pensieve, a recently proposed approach based on
reinforcement learning. The talk is based on work published in NSDI 2017, and
Sigcomm 2018.
Bio: Sanjay G. Rao is an Associate Professor in the School of Electrical
and Computer Engineering at Purdue University, where he leads the
Internet Systems Laboratory. His research spans network verification
and synthesis, low-latency cloud-based web applications, and
Internet video distribution. He received a Ph.D from the School of Computer Science,
Carnegie Mellon University. He has been a Visiting Researcher at Google,
AT&T Research and Princeton University. He is a recipient of the NSF
Career award, and won the ACM SIGMETRICS Test of Time Award for his
work on End System Multicast (peer-to-peer video streaming). He has
served on the Technical Program Committees of conferences including
ACM Sigcomm, Usenix NSDI, and ACM CoNext, has served as the Area
Technical Program Chair of IEEE Infocom, and is currently an Associate
Editor for the IEEE/ACM Transactions on Networking.