|
CS344A is an introduction to systems-oriented research in low-power, wireless
sensor networks. In the context of the course, systems-oriented means
using, discovering, and evaluating approaches which are practical
in real networks. A basic precept of systems work is that there rarely is one
true answer to a problem. Instead, every problem has a set of tradeoffs. Common
examples include flexibility versus safety, performance versus cache size, and
energy versus response time. These tradeoffs mean that which approach is best
in any given situation can vary greatly, depending on the system requirements and
available resources. Often, systems research involves discovering and exploring these
tradeoffs as much as coming up with solutions that fit them.
Wireless sensor networks present a completely different set of constraints than
those which traditional systems grapple with. The most important is energy. These
networks are collections
of devices that must last for a long time on small batteries. Current technology lets
a node running at full power (radio on, CPU active) last for four days; these
four days of activity have to spread across years of life. At these time scales,
costs which most other systems would consider irrelevant -- SRAM maintenance, IO
port leakage -- turn out to be critically important. The second most important
is embedment. Unlike most other computers, sensor nodes are often in
hard-to-reach places, and administrators have very limited visibility into the
operation of the network. Determining whether packet loss is due to an overflowing
queue, dying batteries, or a bird nest can be difficult when the node is 100 miles
away and 30 meters up in a tree. These and additional constraints require novel
systems, protocols, and algorithms.
Because it focuses on the issues and problems that lie between theory and practice,
an important part of CS 344A is providing hands-on experience with low-power
wireless sensor networks. Because a node's radio is its most expensive resource
in terms of energy and also the one it has the least control over due to
environmental variations, many of these problems stem from fundamental challenges
in low-power wireless networking. The major thrust of the class is a final project
that addresses an open research problem. Preparation for the final project involves
reading and discussing research papers as well as two programming assignments which
introduce how to program low-power sensor nodes.
|
