University of Virginia, ESnet Team Up to Win Best Paper Award

May 20, 2013

Contact: Jon Bashor, or 510-486-5849

Networking has several definitions, including making key connections and moving information from one point to another. And a paper co-authored by engineers at ESnet and the University of Virginia proves the value of both -- it was named a Best Paper at the Sixth International Conference on Communication Theory, Reliability, and Quality of Service, held April 21-26 in Venice, Italy.

 The paper, “On How to Provision Quality of Service (QoS) for Large Dataset Transfers,” was written by Zhenzhen Yan, a Ph.D. student at the University of Virginia (UVA), UVA Professor Malathi Veeraraghavan, and Chris Tracy and Chin Guok of ESnet.

 According to Prof. Veeraraghavan, her university had a grant from DOE’s Office of Advanced Scientific Computing Research to study hybrid network traffic engineering. “After hearing a talk by then-ESnet Director Steve Cotter at ESCC 2011, in which he mentioned Chris Tracy's work with Arbor Networks and Packet Design products, we contacted Chris,” said Prof. Veeraraghavan. “I'd known Chris since his days at MAX, when he worked on the NSF DRAGON projects and I was PI of the NSF CHEETAH project.”

 Tracy, Yan and Veeraraghavan started collaborating and in May 2011 submitted a joint UVA/ESnet proposal titled “Terabit-scale hybrid networking,” which was funded by ASCR.

 “We have known Chin Guok for awhile, having met him at DOE workshops and conferences. Chin’s in-depth knowledge of routers and QoS has always been very impressive. So when Zhenzhen started experimenting on the DOE ANI LIMAN testbed we contacted Chin for help with JunOS commands. Chin was very helpful, and provided Zhenzhen many sample scripts. Most importantly, when we showed Chin and Chris our findings from the experiments with the scavenger service setting, Chin suggested the PLP based WRED policing approach, which led to a brand new set of experiments,” Prof. Veeraraghavan. “In summary, Chin and Chris's contributions to this paper were enormously important. Zhenzhen and I are most grateful for their help.”

“A big thanks to Brian Tierney, Eric Pouyoul, Tareq Saif, Chin Guok, Andy Lake, Inder Monga, and DOE ASCR for the 100G testbeds! We also thank Jon Dugan for SNMP data, and Joe Metzger, Joe Burrescia, and Eli Dart for various suggestions. And we thank DOE ASCR program manager, Thomas Ndousse-Fetter for his support of this work.”

Here’s the abstract for the paper: “There is recent interest in using traffic-engineered, QoS-controlled paths for large-sized, high-rate dataset transfers in the scientific community. We refer to TCP flows created by such transfers as alpha flows. Research-and-education network providers are interested in intra-domain traffic engineering systems for identifying alpha flows at ingress routers within their networks, and redirecting them to traffic-engineered paths. This is primarily because of the adverse effects these alpha flows have on delay-sensitive multimedia flows. The focus of this work is to determine what QoS mechanisms are suitable to achieve the dual goals of preventing alpha flows from adversely affecting delay-sensitive flows, while simultaneously allowing them to enjoy high throughput. The interaction between policing schemes on the ingress interfaces and scheduling schemes on the egress interfaces was studied through a set of experiments on a high-speed router testbed. Our conclusions are that a scheduling-only mechanism, with no policing, is well suited to achieve these dual goals if the level of fairness offered by today's IP-routed service is sufficient for simultaneous alpha flows.”