- Speaker: Theophilus Benson (Princeton / Duke)
- Date: December 7th, 2012 (Friday)
- Room: CIT 345
- Title: "Demystifying and controlling the performance of data center networks"
Although there is tremendous interest in designing improved networks for data centers, very little is known about the network-level traffic characteristics of current data centers. In this talk, I will present an empirical study of the network traffic in 10 data centers and show how observations from this study can be used to design more efficient traffic engineering techniques for current data centers. First, I will describe the flow-level and packet-level transmission properties of the applications within these data centers, and their impact on network utilization, link utilization, congestion, and packet drops. I will briefly discuss the implications of the observed traffic patterns for data center internal traffic engineering as well as for recently-proposed architectures for data center networks. Next, I will present an overview of MicroTE, a system that adapts to traffic variations by leveraging the short term and partial predictability of the traffic matrix. I will show that MicroTE performs close to the optimal solution and imposes minimal overhead on the network making it appropriate for current and future data centers. Finally, I will conclude by discussing the implications of my techniques and observations on application performance. Towards that end, I will also give a brief overview of ongoing work to provide better visibility into application performance.
Theo Benson is a Postdoctoral Researcher with Jennifer Rexford’s group at Princeton University. In July 2013, he will be joining Duke University as an Assistant Professor in the Computer Science Department. In July 2012, he completed his PhD in Computer Science at the University of Wisconsin, Madison with Aditya Akella. Before that, he was a back-end software engineer/release engineer for a start-up out of MIT media labs called Open Ratings.