Department for Electrical and Communications Engineering | Networking Laboratory |
S-38.3119 Seminar on Delay-tolerant Networking |
---|
Overview
Today's communication protocols at virtually all layers largely make
a common assumption: end-to-end connectivity exists, forwarding
paths are relatively stable, and communication does not incur
excessive latencies or losses. Transport and application protocols
rely on these assumptions: for example, TCP performance may degrade
quickly with large RTT or repeated non-congestion losses and
application protocols perform poorly or even time out if the RTT
increases significantly or the connectitivity is lost or interrupted
before transactions were completed. Even for dynamic environments
of mobile ad-hoc networking, it is often assumed that end-to-end
paths can be found and that transmission latencies remain at most in
the order seconds.
However, despite striving for ubiquitous connectivity
(always-on or always best connected), we expect poor
connectivity and disconnections to remain for many reasons: a
communication peer or link may not be available, it may not be
economic (i.e., too expensive) to utilize an available link for a
certain task, device constraints (such as low battery) may prevent
communications, or legal regulations (airplane, immigration) or
social conventions (opera, theatre) may prohibit communications for
some period of time. Some communication links may have inherently
long propagation delays and/or may only be available infrequently.
From an application's perspective, in fact, disconnections for some
arbitrary time is nothing else than extra delay so that we can
subsume unexpected disruptions and graceful disconnections under the
same general term of delay.
This seminar on delay-tolerant networking addresses communications
in environments with unusual caracteristics, i.e. properties that
traditional design of communication protocols has not taken into
account. Such "challenged networks" may exhibit, e.g., long
communication delays, unpredictable link availability, and may not
even provide an end-to-end path at all. These characteristics are
partly inherent to certain link layer technologies
(e.g. transmission error rate in wireless networks), but mostly stem
from specific communication settings and system architectures
(e.g. sensor networks connecting underwater equipment or planetary
orbits). We will analyze numerous challenged networking
environments and their communication characteristics. We will
investigate (network,) transport, session, and application layer
solutions to delay-tolerant networking as well as novel networking
architectures dealing with such specific environments and also look
into potential consequences for applications and user interaction
paradigms.
This seminar addresses master and PhD students who (intend to)
specialize in the area of networking, who have already have a sound
understanding of networking protocols and architectures, and who are
interested in detailed analysis, discussion, and further development
of communication protocols.
© 2005 Jörg Ott - Last modified: $Id: index.html,v 1.3 2005/08/14 13:29:25 jo Exp $ |