New Directions in Networked Systems Design - Session 2

Speaker  Emin Gun Sirer, Mohammad Alizadeh, Nathan Farrington, Bruce Maggs, Vyas Sekar, and Brighten Godfrey

Host  Stefan Saroiu

Affiliation  Cornell University, Stanford University, University of California San Diego, Duke University, State University of New York Stony Brook, University of Illinois Urbana

Duration  01:14:52

Date recorded  31 October 2012

Speaker 1 - Copper is Dead — and what that means for computer science

Nathan Farrington, University of California San Diego

Abstract: In this talk, you will learn why copper electrical wiring can never effectively be used to build high-speed data center networks at 40G and up, and no matter how much as computer scientists we kick and scream, we will be forced to learn to stop worrying and love optical communications.

Speaker 2 - Transport in Future Warehouse-Scale Computers

Mohammad Alizadeh, Stanford University

Abstract: The datacenter increasingly resembles a warehouse-scale computer: a large collection of computing and networking resources that work in concert to efficiently deliver superior performance. For this vision to fully materialize, transport in the datacenter (the plumbing) needs to vastly improve. I will discuss some of the opportunities and challenges in this space and recent research that aims to address them.

Speaker 3 - Going from BASE towards ACID with NoSQL

Emin Gun Sirer, Cornell University

Abstract: You've heard of the NoSQL revolution and the CAP Theorem. In this talk, I will tell you why CAP is not what it's cracked up to be and describe a revolutionary new architecture for NoSQL data stores that are consistent, available in the presence of partitions that affect up to a threshold of nodes, scalable, and above all, fast.

Speaker 5 - The Middlebox Manifesto

Prof.Vyas Sekar, State University of New York Stony Brook

Abstract: Middleboxes (firewalls, IDSes, proxies, WAN optimizers, and the like) have long been an integral part of operational networks, but have traditionally been treated as second-class citizens in the research community. There is growing recognition of the need to bridge this disconnect in both camps. I will describe my recent research related to the design, implementation, and management of middle-boxes and discuss some challenges and opportunities in integrating middleboxes with SDN mechanisms

Speaker 6 - Verifying the Data Plane

Prof. Brighten Gdfrey, University of Illinois at Urbana Champaign

Abstract: The increasing complexity of modern computer networks has far outpaced the development of tools to manage their operation. We are developing tools which simplify network security and management by formally reasoning about network-wide forwarding behavior. Our first data plane verification system, Anteater [SIGCOMM'11], revealed multiple real-world bugs in a large university network, including forwarding loops and stale ACL rules. VeriFlow [HotSDN'12] checks network-wide invariants in real time as each forwarding rule is inserted, optionally blocking vulnerabilities from being introduced into the network. Our current OpenFlow-based implementation can perform rigorous checking within hundreds of microseconds per rule insertion. This talk presents work with Ahmed Khurshid, Haohui Mai, Kelvin Zou, Wenxuan Zhou, Rachit Agarwal, Matthew Caesar, and Sam King.

©2012 Microsoft Corporation. All rights reserved.
> New Directions in Networked Systems Design - Session 2