In this project, we explore new protocols, metrics, architectures and applications to improve the efficiency and quality of video delivery on the Internet.


Improving Video Content Delivery

The song "Video Killed the Radio Star" was the first music video shown on MTV and heralded the shift of music and entertainment to the medium of cable television. Video consumption is similarly undergoing a migration from traditional media (e.g., over-the-air broadcast, cable, and satellite TV) to the Internet. Until recently, a U.S. household spends an average of 140 hours consuming video per month, out of which only 9 hours was delivered over the Internet. Within the next few years, a significantly larger fraction of this video consumption will be delivered over the Intenet.

With video taking center stage, these trends create crucial architectural challenges for video delivery and consumption that must be addressed immediately, such as how to provide HD-like streaming quality to consumers, how to accommodate the requirements of heterogeneous clients ranging from smartphones to high-end HDTVs, how to achieve large scale distribution, how to accommodate flash crowds of millions of viewers, and how to satisfy various content and resource related policies, including geographic restrictions on content, and capacity sharing across different content sources such as CDNs and P2P networks. At the same time, the trends represent significant risks for the Internet and related systems, including legacy applications and protocols. Indeed, the popular press has already ``predicted'' potential dire consequences of the predominance of video with headlines such as "Will Netflix Destroy the Internet?".

The primary goal of our research is to address the architectural challenges needed to ensure efficient and optimized video delivery on the Internet given the aforementioned trends, while ensuring no negative impact on the network. Our architecture is centered on an overlay-based control plane for the dominant network traffic type. We believe that this approach will be a fundamental component of any future Internet architecture.

People

Publications

“A Case for Information-Bound Referencing”
by Ashok Anand, Aditya Akella, Vyas Sekar, and Srinivasan Seshan.
In Proceedings of HotNets, (Monterey, CA), Oct. 2010.
Details. Download: PDF.

“Hulu in the neighborhood”
by Dongsu Han, David Andersen, Michael Kaminsky, Dina Papagiannaki, and Srinivasan Seshan.
In Proceedings of COMSNETS, (Bangalore, India), Jan. 2011.
Details. Download: PDF.

“A quest for an internet video quality-of-experience metric”
by Athula Balachandran, Vyas Sekar, Aditya Akella, Srinivasan Seshan, Ion Stoica, and Hui Zhang.
In Proceedings of HotNets, (Redmond, WA), Oct. 2012.
Details. Download: PDF.

“Developing a predictive model of quality of experience for internet video”
by Athula Balachandran, Vyas Sekar, Aditya Akella, Srinivasan Seshan, Ion Stoica, and Hui Zhang.
In Proceedings of the ACM SIGCOMM, (Hong Kong, China), Aug. 2013.
Details. Download: PDF.

“Analyzing the potential benefits of CDN augmentation strategies for internet video workloads”
by Athula Balachandran, Vyas Sekar, Aditya Akella, and Srinivasan Seshan.
In Proceedings of IMC, (Barcelona, Spain), Oct. 2013.
Details. Download: PDF.

“Enhancing video accessibility and availability using information-bound references”
by Ashok Anand, Athula Balachandran, Aditya Akella, Vyas Sekar, and Srinivasan Seshan.
In Proceedings of ACM CoNEXT, (Santa Barbara, CA), Dec. 2013. Best Paper Award.
Details. Download: PDF.

“Modeling web quality-of-experience on cellular networks”
by Athula Balachandran, Vaneet Aggarwal, Emir Halepovic, Jeffrey Pang, Srinivasan Seshan, Shobha Venkataraman, and He Yan.
In Proceedings of ACM MobiCom, (Maui, HI), Sep. 2014, pp. 213-224, ACM.
Details. Download: PDF.

“Practical, Real-time Centralized Control for CDN-based Live Video Delivery”
by Matthew K Mukerjee, David Naylor, Junchen Jiang, Dongsu Han, Srinivasan Seshan, and Hui Zhang.
In Proceedings of the ACM SIGCOMM, (London, UK), Aug. 2015, pp. 311-324, ACM.
Details. Download: PDF.