Ph.D. Thesis Proposal
Open Networking Infrastructure: Boosting Wireless Networks in the Era of Cloud
Tao Jin
College of Computer and Information Science
Northeastern University

In recent years, the mobile Internet underwent revolutionary changes, and has fundamentally changed the way users access the Internet. This resulted in an unprecedented demand for ubiquitous network access. However, due to the fundamental radio frequency communication constraints, the service quality and scalability of cellular systems is limited. Wi-Fi, now being a well-developed standard technology, has been densely deployed, especially in urban area.

My research work is primarily motivated by these exciting trends in mobile networks. We believe WiFi has the unique potential to form a community infrastructure that provides a truly scalable, efficient and ubiquitous access to wireless and data, especially in urban area.

In this work, we propose an Open Infrastructure framework. Based upon this framework, we explore a group of mechanisms to boost the mobile computing experience with the leverage of urban WiFi. To evaluate the feasibility of our ideas, we have built and deployed an Open Infrastructure testbed, consisting of 30 customized home WiFi APs running in the urban areas of Boston and Houston. Since February 2011, we have collected over 70 million residential network usage statistics record and 1.3TB of traffic data trace. The testbed provides us with first-hand information on urban WiFi and a realistic setup to try out a variety of research ideas. Our research has been focusing on two domains, urban WiFi assisted energy saving on mobile devices and idle bandwidth harvesting from home WiFi APs. We have initiated two projects, WiZi-Cloud and BaPu, to develop a set of enabling mechanism and prototypes to demonstrate the feasibility of our ideas.

In WiZi-Cloud, we extend the current WiFi AP and mobile device with an alternative ultra-low power ZigBee radio interface, to reduce the battery consumption on the energy constrained mobile devices. The WiZi architecture and its support for multiple heterogeneous radios is transparent to the applications and allows seamless interfaces switching. WiZi supports multiple ZigBee transceivers and channel coding schemes. We design and prototype a complete suite of hardware/software solution. Our experimental results show that the WiZi-Cloud well supports a large set of mainstream mobile applications and improves the energy efficiency by 3 folds and can exceed WiFi coverage range.

BaPu is motivated by the fact that today's residential broadband connections have limited backhaul bandwidth, especially in uplink, which highly constrains many fastly growing applications, such as HD content instant sharing, efficient cloud storage backup, etc. In BaPu, we design the mechanisms of aggregating the idle broadband uplinks, by having the mobile device communicate with multiple proximate WiFi APs in the same neighbourhood. BaPu is a complete software solution running on home WiFi APs, and requires no modifications to the client devices. This makes an easy incremental adoption of BaPu technology. Our architecture and protocols design provide a clean solution to several challenges in efficiently supporting both TCP and UDP. Our prototype system shows that BaPu can efficiently aggregate the backhaul bandwidth of multiple access points and achieves up to 90% of the theoretical maximum throughput.

Download: [PDF]
Slides: [PDF]
Thesis Committee
Guevara Noubir (Advisor)
Prasant Mohapatra (External Member, University of California, Davis) [CV]
Justification for thesis committee composition:
Professor Mislove is an expert in networked systems. Professor Rajaraman is an expert in network algorithms in general and wireless networks algorithms in particular. Professor Mohapatra expertise spans Wireless LANs, Wireless Mesh Networks, Wireless Sensor Networks, Overlay Networks, Internet Measurements, and QoS provisioning in wired and wireless networks ( The committee members provide a good mix of expertise to evaluate the thesis work.
Tao Jin, College of Computer and Information Science, Northeastern University, Boston, MA