Spotlights in Our Work

We are sharing more details about a few research projects, on which we spent a lot of time and efforts in the past decade.

In the Spotlights column, we profile some of our past and ongoing research projects, each in their dedicated pages. We strive to make them easy to read and to write in layperson’s terms, rather than just a cut and paste from the abstracts of research papers. We may include presentation slides in Adobe PDF and Flash slide shows to make it a bit easier to go deeper into the topic. We would still sometimes link to the related papers, though, just for easier references.

Agilos (1996 — 2000): Quality of Service adaptations made easy. This spotlight feature has been authored in December 2006, in dedication to the ten-year anniversary of the project launch in summer 1996.

Magellan (2006 — 2008): Live peer-to-peer (P2P) multimedia streaming applications have been successfully and commercially deployed in the Internet with up to millions of users at any given time. Wouldn’t it be great if we can obtain a complete and detailed understanding of the topologies, performance metrics, peer dynamics, and ISP clustering effects of such large-scale real-world streaming systems? With the Magellan project, we collaborate with UUSee Inc. — a leading P2P streaming solutions provider in China — to achieve these goals. We attempt to add instrumentation code directly into the commercial product, rather than treating it as a black box, as most other existing studies did.

Network coding implementations (2007 — 2010): How practical is network coding? How to implement network coding in the most efficient fashion?

Streaming with Network coding (2007 — 2010): Is it realistic to expect that network coding can be successfully applied in real-world peer-to-peer streaming systems? How do we design new protocols and algorithms to take full advantage of network coding? What may be the most prominent advantages and drawbacks of network coding in streaming systems?

Stable matching in networking (2010 to present): Can the classical stable matching theory from economics, as well as the elegant deferred acceptance algorithm extensively used in real-world markets, provide insights and inspire simple and practical solutions for networking problems that exhibit a market-like setting? What are its edges and downsides compared to conventional solutions, such as optimization?

GestureFlow: Streaming Gestures to an Audience (2010 to present): Can multi-touch gestures be streamed in a collaborative session with multiple participants? What are the requirements in such a session that streams multi-touch gestures?

Network Coding for P2P Content Distribution (2007 to present): How to theoretically assess the benefit of network coding in P2P content dissemination? Can random network coding simplify the neighbor selection and rate allocation algorithms that are conventionally carefully designed to achieve the optimal download rate? How do topological effects affect network coding, and what can network coding offer beyond throughput benefits?

Multi-Resource Packet Processing (2012 to present): Middleboxes are widely deployed in today’s networks. They apply a variety of complex network functions to transform, filter, and optimize incoming traffic based on the payload of packets. These functions require the support of multiple types of resources, such as CPU and link bandwidth, for processing incoming packets.  How can we design a multi-resource packet scheduling algorithm to allow flows to share these resources fairly and efficiently?

Bandwidth Allocation in Datacenter Networks (2013 — present): Large scale datacenters have become the de facto standard computing platform for data parallel applications, which typically have their tasks distributed across servers and demand network bandwidth for communication among tasks. As the datacenter network is globally shared by multiple applications, how should we fairly allocate link bandwidth among them? How can we optimize the network design to improve application performance?