Guest lecture: Marco Chiesa "Terabit-per-second Programmable Networks"

Terabit-per-second Programmable Networks

Marco Chiesa
KTH, Sweden

Abstract: Flexibility and performance are commonly considered two contrasting requirements in today's networks. Flexibility allows operators to deploy and move network functionality anywhere in their network, thus supporting efficient resource utilization. Deploying a general-purpose yet power-hungry infrastructure is a common way to enable flexible network control. Conversely, performance require specialized chips that are tailored for a narrow set of possible computations. The current sheer growth of both Internet traffic and network-related energy consumption begs for new approaches to networking that can recombine the benefits of flexible network control with the performance of fixed-function networks.

In this talk, I will present our journey toward designing packet processing architectures that are both flexible and performant. I will set forth my view on how future networks will heavily rely on network disaggregation principles to carefully separate responsibilities between different network entities based on their unique constraints. I will ultimately show that it is today possible to process the equivalent of one terabit of traffic on a single CPU socket. Finally, I will also provide general insights and future directions for programmable networks, including implications of our work to compiler and hardware design.

Short bio: Marco Chiesa is an Associate Professor in the Network System Lab group of the Software and Computer Systems division at the KTH Royal Institute 
of Technology, Sweden. His research interests lie in computer networking and, more specifically, in aspects of Internet protocols and architectures ranging from security to network optimization, including Software Defined Networking (SDN) approaches to these challenges, next-generation Internet eXchange Points (IXPs), and beyond. He is the recipient of the Usenix NSDI Community, IEEE William R. Bennett, and IETF Applied Research Network awards.