Delay Tolerant and Opportunistic Networks

Abstract

We  are so used to networks that are "always there", so called
infrastructural networks such as the phone system, Internet, the cellular (GSM,
CDMA, 3G) and so on that we forget that once upon a time (why, only in
the 1970s) computer communications was fraught with problems of
reliability, and challenged by very high cost (or availability) of
connectivity and capacity. One we had UUCP and E-Mail, which predate
any of today's infrastructures, but coped very well with these
challenges. Now, it appears that it is worth revisiting these ideas
for a variety of reasons: it looks like we cannot afford to build a
Solar System wide Internet just yet; it looks like one can build
effective end-to-end mobile applications out of wireless
communication opportunities that arise out of infrequent and short contacts 
between devices carried by people in close proximity, and then wait til these 
people move on geographically
to the next hop; it is interesting to speculate that these systems may
actually have much higher potential capacity than infrastructural
wireless access networks, although they present other challenges
(notably higher delay). This set of talks will be about the last 10
years of work leading up to our current understanding of how to build
Delay Tolerant and Opportunistic Networks, and how to model their
performance.

Lecture 1

Delay Tolerant Networking - it really is rocket science.

In this lecture we review the DTN work over the last 10 years,
starting from the origins as an initiative to provide a commodity
network system for the planned manned mission to Mars,  and ending up
with an architecture for any network that is challenged by frequent
disruptions, from oceanographic sensing, to disaster relief
communications when everything else has failed, and on to building
useful services in places where the population cannot (yet) afford an
infrastructure.


Lecture 2

Opportunistic Networking - Making people network.

4 billion people have cell phones. Most have not just a radio for
voice communication, but also a bluetooth, short range radio which
allows devices to communicate directly (without involving a cell
tower). Increasingly, devices also have WiFi which can also be used
without any recourse to a provider. We can build networks that use 
encounters between devices carried by people, and then use the natural
mobility of humans (walking, cycling, driving, in trains, planes etc)
to carry stored data to the next hop. Such systems can be used in a
wide variety of scenarios for disaster communication when the
infrastructure is broke, for networks in developing regions (or out in
the middle of the ocean, or in space) where
there isn't any infrastructure anyway, and for applications which may
enjoy high capacity, but do not mind higher (or even uncertain)
delays. In the process of designing and building such systems, we may
accidentally (on purpose) design systems tat simply work better in the
now more  traditional setting of the Internet, but cope more
seamlessly with the occasional glitches that show up there. We may
find it easier to build applications on such systems that tolerate
occasional (or frequent) disruptions. One interesting synergy I will
touch on here is that applications for opportunistic networks often
entail unspecified sender or recipient (i.e. they are data
dissemination applications) which resemble some of the new ideas in
Data Oriented networking in the Internet. At the same time as resource
pooling and multipath routing, and interest-based delivery are being explored for the Internet to
support this, such approaches have already proved natural in designing
data forwarding schemes in Opportunistic Networks. We'll look at one
such protocol.

Lecture 3 

How much delay must I tolerate in my DTN/Oppnet?

It turns out that we have a wealth of data emerging from measurement
made by wireless networking researchers (e.g. on Crawdad) but also
increasingly of interest to social scientists (anthropologists trying
to understand human society) and medical researchers (epidemiologists
trying to understand the spread and evolution of diseases). 
In this final part of the lecture, I will look at the emerging models 
we have both for delivery success and delay, and for capacity of DTNs.
