The Internet under widespread shelter-in-place: Resilience, response, and lessons for the future
PIs: Matt Calder, Dave Choffnes, Ethan Katz-Bassett
The spread of COVID-19 has led to unprecedented and ongoing changes to daily life, including shelter-in-place orders, widespread closing of businesses and schools, and work-from-home and school-from-home at previously unknown levels. These changes in behavior are placing extraordinary demands on the Internet. This project will measure the Internet’s ability to meet these demands, including comparing its performance before, during, and after the peak of COVID-19; whether the amount of change varies between areas heavily impacted by COVID-19 and those less impacted; and whether and how large networks adapt.
To provide this rich understanding, this project will combine multiple Internet-scale datasets that provide complementary views to investigate how responses to COVID-19 have impacted the Internet and how networks have reacted. Specifically, the project will measure how workloads have changed and their impact on networks in response to COVID-19 by measuring traffic patterns worldwide and the impact of congestion on performance globally. In addition, the project will measure how networks react to these changes in workloads by monitoring path changes that indicate rerouting of traffic during surges induced by measures in response to COVID-19, identifying whether traffic management is deployed for (de)prioritizing certain network flows, and measuring whether networks add interconnections and upgrade capacity.
Measuring the network impact of COVID-19 will illuminate the Internet’s strengths and weak points and is a crucial step towards improving the Internet’s future resilience in the face of pandemics, natural disasters, large scale conflict, and terrorist attacks. The Internet is designed for resiliency, but large cloud and content providers have been avoiding public Internet exchange points and the public Internet in favor of private interconnection and private wide-area networks. This project will understand the impact of these decisions. The relative performance of these strategies under stress should play a critical role in informing future research, funding, and design of critical Internet infrastructure.
All of the data, measurement tools, and analysis code used for this study will be published to the greatest extent permitted by our affiliated organizations and prevailing data privacy regulations, with links to these artifacts posted at http://covid19-internet-resilience.github.io/. The goal is to keep these artifacts publicly available in perpetuity via free storage and cloud repositories, though the duration is subject to the availability of such resources. The tools and code will be maintained for as long as they remain useful for measuring Internet resilience.