[Nrg-l] PhD Oral Area Exam of Karim Mattar -- Mon 12/18 @ 4pm
matta at cs.bu.edu
Sat Dec 16 10:08:09 EST 2006
PHD ORAL AREA EXAM
Computer Science Department
PhD Candidate: Karim Mattar
Date: Monday, December 18
Place: Room MCS 135, 111 Cummington Street
(for directions, see www.cs.bu.edu/colloquium)
Title: Cellular Data Networks: Integration Problems,
Solutions, and Future Directions
Internet users today expect global connectivity anytime from anywhere in
the world. This anywhere-anytime idealogy for Internet access, despite
being grandiose, requires a tremendous infrastructure to provide the
required coverage. Cellular networks do possess such an infrastructure
making their integration with IP a natural next step in the evolution of
the Internet. The result was the birth of cellular data networks---a
hybrid system that poses new networking problems and challenges.
In this talk, I primarily survey recent work dealing with two aspects of
a cellular data network, namely: a) the wireless resource allocation
problem, focusing on solutions to fairly and efficiently share a
time-slotted wireless channel among many users, and b) transport layer
solutions that are optimized for such wireless channels.
With regard to the transport layer, I classify the solutions based on
their primary mechanisms, which are designed to alleviate the effect of:
a) wireless losses, b) mobility/handoff, c) bandwidth asymmetry, and d)
variable rate/delay links. With regard to the wireless resource
allocation problem, I identify three classes of scheduling models, each
with its own assumptions, under which solutions were developed, namely:
a) queue-centric, b) queue-oblivious, and c) adversarial models.
I will argue that none of the existing transport layer solutions are
adequate for time-slotted wireless channels that employ sophisticated
channel-aware schedulers and are notorious for having highly variable
rates and delays. I will also discuss the possibility of developing
transport layer solutions that specifically deal with adversarial
channel models that capture the effect of worst-case channel behaviors.
The exposition of the resource allocation and transport problems
naturally leads to discussing potential directions for composing the two
components safely and efficiently.
More information about the Nrg-l