Recognizing the importance of SDN issues, over the last few years I have set up a lab to enable my research students to work in this area. Since new topics are slow to enter the curriculum, I have also attempted to design the lab to enable instructional use. By the kind cooperation of the Computer Science department, the lab was initially housed in the NetLabs area in EB2-East, the physical equipment has now been moved to my own lab in EB2-B029, and it is remotely accessible for students. It includes networking equipment donated by various industry partners, purchased from grants, and also surplus computing equipment obtained from NetLabs. See the main SDN Labs website for more information.
Many emerging security and network research questions are in areas of network availability, reliability etc. Solutions are often proposed in research through routing, opportunistic MAC, adaptive power control, dynamic rate control and modulation, and other such research areas; these are all low down, or across traditional boundaries of, the networking stack. Commercially available wireless networking equipment do not allow experimentation at these detailed levels, and researchers are reduced to using commodity computing hardware and using them as wireless equipment.
We envisioned CentMesh, and subsequently architected and developed it,
to provide a versatile wireless networking substrate that would be deeply
programmable (to allow whatever research innovation it was called upon to
support), but provide a flexible and modular interface (to allow a researcher
to make specific changes in the programming relevant to the research without
requiring to undertake a large software project). Further, it is
architected to be extensible; the basis of ever more ambitious and powerful
research enabling infrastructures to come, as researchers pursue evolving
research directions, both guiding and contributing to the enhancement of
CentMesh capabilities. CentMesh is thus a research project that provides
usable facilities, not a development project that produces a static platform.
More information can be found at the CentMesh website. CentMesh was covered by The News and Observer - the daily paper of the Raleigh-Durham-Chapel Hill area. One brief clarification to the title of the archived news story (which attempts to capture the spirit of the story while staying within the brevity required of newspaper titles). CentMesh is a first in many ways, but it is obviously not the first outdoor Wi-Fi network. There are many commercial networks which do that, including the one that covers Raleigh downtown. There are also experimental outdoor WiFi testbeds. However, the combination of programmability, span, and complete researcher control of CentMesh is indeed unprecedented as far as we know.
By definition, this section is perpetually under construction. Depending on when you came here, there might not be much here; hopefully there will be soon, check back.
Not all the research done by all my students is represented in here - not all thesis work is a "project" or part of one. That is why we have dissertations and publications - which after all are more important than my arbitrary organization here.
As with all research, these projects are dormant rather than ended - any of these areas could be topics for further research at a future time.
To build this unprecedented facility, NSF funded BBN to set up the GENI Project Office (GPO), which in turn funded researcher groups to architect and develop parts of the whole. My Integrated Measurement Framework (IMF) project is one of these - the only one at NCSU so far that has been funded by GENI. Building GENI is an engineering exercise, but one that demands that experience and awareness in breadth of networking research inform it at every point, which is why GPO is funding various researchers and research groups to build GENI, rather than doing it all at BBN. GENI follows the processes and standards of deliverables that are more typical of production-grade engineering.
More information about the actual project can be found at the IMF project wiki at GENI. In this last year, we are designing the Measurement Plane messaging system for the instrumentation and measurement cluster of GENI. This is forming a critical piece of both GEMINI and GIMI - the two redundant Instrumentation and Measurement architectures being developed for GENI (I am part of the GIMI effort).
More information is at the SILO project website. The project ended after two years, with several publications as well as a working prototype proving the concept of our paradigm, and an REU supplement. However, the concepts we developed in SILO continued to influence our Internet architectural work - we have applied them both in the GENI-IMF project and ChoiceNet.
The lablet is pursuing various units of research aimed at security science topics, led by various researchers. I am engaged in two such units. The first is a project that addresses a specific collaborative approach for Sybil detection, suitable for sensor networks, that one of my students had previously worked upon. This project will investigate the limits of applicability of such an algorithm, and obtain detection probabilities under various conditions. The second project I am leading is on attempting a scientific description of network security mechanisms as control systems. This project will investigate specific network security systems by viewing them as feedback control systems, and attempting to determine general stability results or other characterization for them. Dr. Meeko Oishi from University of New Mexico is a co-PI on this project.
The quantum jump in transmission speed due to the use of fiber has not been
matched by a similar improvement in processing speed at intermediate
network processing elements (routers, switches), and as computational speeds
go up, so does throughput. As a consequence, either the
fiber has to be severely underutilized (not a preferable solution, and not
viable in the long run), or many processing elements have to be deployed to
conventionally route/switch at each intermediate (network interior) node. This
latter solution can be prohibitively expensive. In addition, the processing
equipment is all electronic, thus delay is incurred in electro-optic conversion
every time the packets/cells contained in an optical signal must be routed.
This will usually be necessary because the bandwidth of even a single
wavelength is likely to be much larger than the typcial user channel bandwidth,
thus many slower speed traffic streams will be multiplexed (probably TDM) over
each wavelength channel. It becomes very attractive to allow some (hopefully
the bulk) of the traffic to be switched optically, using wavelength routing,
and resort to electro-optic conversion and electronic processing only when it
cannot be avoided. This problem has been called traffic grooming in literature.
In the general case, as well as in many quite restrictive cases of topology and
traffic patterns, this problem is computationally intractable.
Over the years, I have worked with many students on various topics related to grooming. Some of it was archived at the Traffic Grooming website, but unfortunately that website has fallen out-of-date. Please see my 2007 and 2002 surveys on grooming, and also the book I co-edited in 2008.
This project was jointly conducted with Dr. Mihail Sichitiu of the ECE
department and Dr. Sami Rizkalla of the CCEE department of NCSU. The goal of
this project was to increase the lifetime of a battery-powered sensor network
using a combination of scheduling and power aware routing for continuous
monitoring sensor networks. The motivating application is structural health
monitoring of building and bridges.
We developed algorithms to schedule sleep cycles of the sensors, to improve the lifetime of the network, to redistribute the energy of the network for maximum utilization, and to perform power-efficient routing. The final and highly relevant step for this project was the practical implementation of a signal pre-conditioning circuit featuring a programmable amplifier capable of variable range and resolution as well as temperature and non-linearity compensation. This was tested with actual bridge beams at the Constructed Facilities Laboratory at NCSU, with readings received remotely. The complete project is archived at the WALAN website.