Friday, November 8, 2019
Towards Highly-Available Cloud and Content Provider Networks
Ramesh Govindan, University of Southern California
11:00am, DBH 6011 [Host: Nalini]
Large content and cloud providers like Google, Facebook, and Microsoft are increasingly pushing for high availability while still attempting to ensure low latency for their customers and high utilization of their infrastructure. Achieving these goals, while simultaneously permitting high-velocity evolution of their infrastructure to support growing demand and new services, is a significant research challenge.
In this talk, I will draw upon two recent examples of work from our group that highlight some of these challenges, and suggest potential solutions. The first work explores failure masking in content provider wide-area network routers. These routers have novel interconnect structures that can be exploited, with clever internal traffic routing schemes, to mask a large number of internal failures. The second work examines an important aspect of availability loss in modern networks: network management operations that are needed to evolve the infrastructure. In this work, we explore the question of whether it is possible to, and what it means to, design networks for manageability.
Ramesh Govindan is the Northrop Grumman Chair in Engineering and Professor of Computer Science and Electrical Engineering at the University of Southern California. He received his B. Tech. degree from the Indian Institute of Technology at Madras, and his M.S. and Ph.D. degrees from the University of California at Berkeley. His research interests include routing and measurements in large internets, networked sensing systems, and mobile computing systems.