Email: tbenson@cs.duke.edu
Phone: (919) 660-6555
Skype: theobenson
D342 LSRC Building
Durham, NC 27708

About Me

I am an Assistant Professor in the Computer Science Department of Duke University. My research interests include solving practical networking and systems problems, with a focus on Software Defined Networking, data centers, clouds, and configuration management. In the past, I've conducted large scale measurement studies of data centers and enterprise networks; and I've developed several networked and distributed systems -- one of which was purchased in 2012. To date, my study on data center traffic characteristics has been cited in over 800 papers.

News

Selected publications (full list)


Downloads: PDF
The Case for Making Tight Control Plane Latency Guarantees in SDN Switches.
Huan Chen and Theophilus Benson
Proceedings of SoSR 2017, Santa Clara, CA, April 2017]


Downloads: PDF
Dapper: Data Plane Performance Diagnosis of TCP.
Mojgan Ghasemi, Theophilus Benson and Jennifer Rexford.
Proceedings of SoSR 2017, Santa Clara, CA, April 2017]


Downloads: PDF
A View from the Other Side: Understanding Mobile Phone Characteristics in the Developing World.
Sohaib Ahmad, Abdul Lateef Haamid, Zafar Ayyub Qazi, Zhenyu Zhou, Theophilus Benson and Ihsan Ayyub Qazi.
Proceedings of IMC 2016, Santa Monica, CA, Nov 2016]


Downloads: PDF
Performance Characterization of a Commercial Video Streaming Service.
Mojgan Ghasemi, Partha Kanuparthy, Ahmed Mansy, Theophilus Benson and Jennifer Rexford.
[Proceedings of IMC 2016, Santa Monica, CA, Nov 2016]


Downloads: PDF
Tolerating SDN Application Failures with LegoSDN.
Balakrishnan Chandrasekaran, Brendan Tschaen, and Theophilus Benson.
[Proceedings of SoSR 2016, California, US, Mar 2016]
A workshop version appeared in HotNets 2014.


Downloads: PDF
Picocenter: Supporting long-lived, mostly-idle applications in cloud environments.
Liang Zhang, James Litton, Frank Cangialosi, Theophilus Benson, Dave Levin, and Alan Mislove
[Proceedings of EuroSys 2016, London, UK, Apr 2016]


Downloads: PDF
Programming Slick Network Functions.
Bilal Anwer, Theophilus Benson, Nick Feamster, and Dave Levin.
[Proceedings of SOSR 2015, Santa Clara, CA, Jun 2015]


Downloads: PDF
Demystifying Configuration Challenges and Trade-offs in Network-based ISP Services.
Theophilus Benson, Aditya Akella and Aman Shaikh.
[Proceedings of SIGCOMM, Toronto, Canada, Aug 2011]


Best Paper
Downloads: PDF
Network Traffic Characteristics of Data Centers in the Wild.
Theophilus Benson, Aditya Akella and David Maltz
[Proceedings of IMC, Melbourne, Australia, Nov 2010]


Projects

  • Fault Tolerant Software Defined Networks: SDN controllers are poorly architected and error-prone, yet, availability of modern networks are inexplicably tied to these controllers. In our work, we take a two-pronged approach to improving SDN availability. First, we improve the fault tolerance of SDNs by rearchitecting SDN controllers to include primitives that engender higher reliability and minimize downtimes. Second, we leverage the insight that complicated bugs manifest themselves in production networks and develop a set of frameworks and algorithms to safely and proactively detect bugs in production environments.
  • Abstractions for Software Defined Infrastructures: Software defined infrastructures, e.g. clouds and cloudlets, are being designed and deployed to support a variety of emerging applications and use cases, e.g. big data analytics and IoT devices. Unfortunately, SDIs lack appropriate abstractions to support these emerging uses cases. Furthermore, the topology and networks of many SDIs are developed using cost-inefficient algorithms. To this end, we are currently engaged in several projects to develop such abstractions and algorithms, including; abstractions for supporting finer-grained computation in clouds (or cloudlets); abstractions for dynamically orchestrating virtual network services (virtualized network functions); and, algorithms for efficiently designing and routing on cost-efficient topologies.
  • Software Defined User Experience: Networking protocols are designed to be broadly applicable to a wide range of scenarios. This generality makes the protocols ill-suited for specific situations and ill-suited to enforce a user's rich and dynamic expectations of performance. For example, networking protocols designed for first world have proved ineffective when run in developing countries and networked applications aim to deliver equivalence quality regardless of the situation and the user's expectation (aiming for 1080p video on a mobile phone -- when lower bits rates will be equally as suitable).
    To address this, we are working on developing (1) frameworks that detect the contexts in which applications and devices are being used and (2) algorithms that make resource allocations or configuration changes that optimize end-user performance. Specifically, we have projects: to detect user's expectations of performance under different contexts and by adjust allocations of resources in home networks to align with these contexts; to design performance-oriented network services that optimize networking protocols to features of a device's context (networking environment, Operating system, and networking stack); and to develop systems for enabling performance-oriented network services to interoperate on secure web pages.

Funding: Grants & Gifts

  • NSF Medium:Towards Finer-Grained Cloud Computing, 2014 - 2017.
  • Facebook Faculty Award, 2015, 2016.
  • Google Cloud Resource Usage Grant, 2015.
  • Yahoo! FREP, 2014, 2015.
  • Yahoo! ACE, 2014.
  • 2014 Equipment Gift from Yahoo, 300 servers.
  • Facebook Gift, 2014-2016.
  • Duke Bass Connections, 2014.

Copyright © Theophilus Benson 2016 · css design: Thomas Park