Devavrat Shah and Damon Wischik.
Abstract.
Much research in networking, in areas such as TCP congestion
control, wireless MAC protocols, fair sharing of data
centers, and optimization of distributed algorithms, falls under
the general heading of resource allocation. The network
modeling community has developed a general-purpose
language for describing resource allocation problems, and
canonical mechanisms for solving them. We believe these
should be taught as fundamental principles to graduate students
in networking. This will encourage themto apply ideas
from one area of networking to another, it will help them to
distinguish which parts of a problemneed hands-on systemslevel
work and which do not, and it will make them think
more deeply about economic and social questions such as
network neutrality.
The first step in any resource allocation problem is to
decide what the capacity region is. In this example there are
two wireless transmitters and two receivers; each transmitter
has a maximum power, and the transmitters can adjust their
power.
What throughputs y1
and y2 can be achieved,
assuming the transmitters set their power levels optimally?
How does the answer change if the receivers try to decode
both signals, rather than just their intended signals?