A Primer to My Dissertation Research

Rohan Narayana Murty


Opportunistic Wireless Networks

School of Engineering and Applied Sciences

Harvard University (2011)

My dissertation research

In my dissertation research, I develop techniques to improve wireless network efficiency and capacity via opportunistic wireless network architectures.


My dissertation research is broadly motivated by two prevalent trends:

1. With an increase in the proliferation of various forms of mobile devices, wireless networks are slated to become the dominant method of network access of the future. Such advances have begun to place a premium on radio spectrum, which is fast becoming a scarce and expensive resource. However, despite the significant growing pressures on spectrum in current wireless networks, there exist large portions of the overall spectrum that are severely under-utilized.

2. Most wireless networks have been designed to operate within statically allocated slabs of spectrum. This approach restricts the capacity of wireless networks even when there exist opportunities to better use the spectrum. For example, even if other portions of spectrum lie unused, Wi-Fi devices must only use the allocated 2.4 GHz or 5 GHz bands.

What implications do these trends have?

The spectrum commons theory considers spectrum to be a good owned by the people and not by any private entity or the government. But entities may be given licenses to use portions of the spectrum exclusively for limited periods of time. Hence, when viewed under the lens of this theory, these two observations point to a course of action: when a slab of spectrum is not being actively used by the licensee (or incumbent), it must be available for use by another entity, insofar as there is no hindrance to the incumbent (original licensee).

Driven by these observations, my dissertation work has developed the principles and techniques to build opportunistic wireless networks, which work by continually seeking and using portions of spectrum currently underused by incumbents, and being agile enough to vacate portions of the spectrum if any incumbent returns.

Where might this work be applicable today?

A prominent emerging system where opportunistic wireless networking can work well is in the so-called white spaces. White spaces are those channels that, in an instant in time, are not used by the incumbents: television stations or wireless microphones. The historic decision by the Federal Communication Commission (FCC) in 2008 permits the operation of unlicensed devices over these white spaces, as long as such operation does not hamper incumbent operation.

WhiteFI: White Space Networking with Wi-Fi like Connectivity

[PDF] [Slides]

WhiteFi demonstrates the need for new approaches in designing algorithms and protocols when networking over the white spaces. The key observation behind this work lies in the characterization of white spaces spectrum as being fragmented and varying spatially and temporally. Hence, it makes the case for why conventional Wi-Fi based techniques do not suffice. Consequently, WhiteFi proposes new adaptive algorithms and protocols for spectrum assignment, association, and handoffs, which factor in the characteristics of the white spaces spectrum. A prototype implementation and evaluation of WhiteFi reveals the need and efficacy of these new approaches in enabling opportunistic networking over the white spaces. WhiteFi takes the first stab at networking over the white spaces and in doing so, it lays the ground work for future work that builds on it.

Dyson: An Architecture for Extensible Wireless LANs


Dyson extends the SenseLess architecture to enable a programmable wireless network. This permits site-specific customization of the wireless network and the ability to quickly roll out innovations into the wireless network. Conventional wireless network designs, often stifled by the pace of standardization, are not architected to embrace such changes easily. Dyson defines a set of APIs that allow clients and base stations to send pertinent information such as radio channel conditions to a central authority such as the SenseLess database. The central authority processes this information to form a global view of the network and make decisions. Dyson provides a Python-based scripting API that allows the central authority’s policies to be extended for site-specific customizations and new optimizations that leverage historical knowledge. A prototype implementation of Dyson reveals the ease with which network functionality can be changed.

The initial vision for Dyson appeared in our HotNets paper as Trantor

What is the roadmap for my dissertation work?

My dissertation work primarily consists of four major projects. Briefly, WhiteFi (SIGCOMM 2009, best paper award) is the first Wi-Fi-like network to function over the white spaces. SenseLess (DySpan 2011) eliminates the need for spectrum sensing, a key stumbling block in conventional white spaces networks. Dyson (USENIX ATC 2010) builds on SenseLess to develop an extensible wireless network platform that permits administrators to perform site- specific customization of the wireless network. DenseAP (NSDI 2008) demonstrates the efficacy of a Dyson-like architecture in exploiting wireless deployment density via opportunistic policies to increase network capacity.

SenseLess: A Database-Driven White Spaces Network

[PDF] (Under Submission)

SenseLess does away with the need for spectrum sensing as a means to determine spectrum availability. Spectrum sensing is hard to achieve reliably at signal low thresholds, thus resulting in conservative policies on spectrum use. In SenseLess, in a significant departure from conventional wireless networks, nodes rely on a database to determine spectrum availability. In this context SenseLess proposes new algorithms and protocols for networking including discovery and handling mobility. It uses a combination of an up-to-date database of incumbents and white space devices, sophisticated signal propagation modeling, and an efficient content dissemination mechanism to ensure efficient, scalable, and safe white space network operation. Comparisons between a prototype SenseLess implementation and ground truth spectrum measurements performed across a 1500 mile path reveal the system’s efficacy in protecting the incumbents while permitting white space use. It is the first instantiation of a white spaces network operated entirely using a database and has also been part of the effort to demonstrate to the FCC of the benefits of obviating sensing.



DenseAP leverages a Dyson-like architecture to exploit increased density of wireless deployments to improve the overall network capacity. This is achieved by an adaptive spectrum assignment algorithm that is tightly coupled with a dynamic association protocol. DenseAP does not require any changes to legacy devices, thereby making it easier to deploy and adopt. An evaluation of a prototype implementation of this system reveals orders-of-magnitude improvement in overall network capacity.

The techniques espoused by my dissertation work, though primarily demonstrated over the white spaces, extend to any wireless network where devices access spectrum opportunistically. A common thread across my solutions to problems is a strong desire to demonstrate the power of an idea by building, deploying, measuring, and analyzing it. When applicable, I have relied on theory and simulations to bolster the case for the systems I build. I have also used tools such as optimization, information theory, and game theory in my work.

How were these ideas demonstrated?

To prototype these ideas, I have deployed a host of wireless testbeds, most of which have been opened up for other researchers to use. I have deployed a floor wide white spaces testbed consisting of USRP nodes backed by an implementation of the SenseLess database. A SenseLess implementation has also been made available for public consumption here. The service exposes a SOAP-based API interface for researchers to use the database for their own white spaces research. I have been part of a team deploying CitySense, an urban outdoor wireless mesh testbed across the rooftops of Cambridge. I have also deployed two separate indoor floor wide testbeds of Wi-Fi nodes.

Was it all computer science?

Overall, my dissertation problem lends itself naturally to collaboration across different domains. To enable such opportunism in wireless networks, I have had to move beyond the concerns of architecting systems in a vacuum. In addition to the technical challenges in the network stack, there is an interplay between them and regulatory, policy, and economic concerns. I have understood the technical details, the regulatory concerns, and economic incentives when architecting opportunistic wireless networks. Consequently, I have collaborated with researchers in information theory, digital signal processing, mechanism design, programming languages, and law.