Transcript Presentation

DTTP: A Delay-Tolerant Transport Protocol
for Space Internetworks
Christos Samaras
ComNet Group, Democritus University of Thrace
February 2008
Contents
1.
Space Networking Environments
2.
Standard Internet Protocols in Space
3.
Space Agencies vs. Each Other
4.
DTTP: Delay-Tolerant Transport Protocol
5.
Simulation Results and Future Work
Space Networking Environments
Challenged networks:
•
intermittent connectivity
•
long and/or variable propagation delays
•
asymmetric data rates
•
high error rates
So, can standard Internet protocols operate in Space?
Standard Internet Protocols in Space?
short answer
no (at least in their current form)
long answer
maybe (adaptations needed: Mobile IP? TCP not suitable; etc.)
Internet usability is based on the following assumptions:
•
continues, bidirectional end-to-end path
•
short round trip times
•
symmetric data rates
•
low error rates
Space Agencies vs. Each Other
Space missions interoperability:
– common goal for many space agencies (CCSDS efforts)
– increase in data return rates
– offering flexible/alternative communication opportunities
– might prove catalytic in critical situations
...no consensus (yet) among space agencies
potential space communications convergence through:
•
deploying common protocol stacks (possibly IP-enabled)
•
hiding heterogeneous networks (e.g. Delay-Tolerant Networking (DTN)
architecture as a message-oriented overlay)
•
other (to be conceived)...
in any case, we need a
specialized, efficient, reliable transport protocol
Why a Transport Layer Approach?
•
ease-of-use: programmers are familiar with developing applications which
sit upon a transport layer
•
the DTN approach only disguises congestion; need for mechanisms that
handle congestion or storage capacity depletion
•
there are cases where homogeneous networks (in terms of underlying
protocol stacks) don’t require different DTN protocols for each hop: a
multi-hop transport solution is therefore needed
DTTP: Delay-Tolerant Transport Protocol
DTTP features:
•
reliability: asynchronous acknowldgement procedures (when compared to
TCP’s Ack-clocking functionality)
•
custody transfer: based on in-network storage; robust against link
disconnections; more efficient than end-to-end approaches
•
parallel data transfer: multiple data paths can be exploited in parallel
•
(time periods with) constant sending rate: rated-based protocol; fills the
communication pipe (note: stateful sessions)
•
sending rate adaptivity: relies on explicit signals from (intermediate/final)
receivers, e.g. storage exhaustion
•
application-oriented transmission behavior: provision of transmission tactics
to reflect application needs
1st Transmission Tactic
immediate use of acknowledgment info;
graduated reliability enhancements (help: redundant data);
suitable for certain video or image applications etc.
until (all application data is acknowledged)
start transmitting new application data
if (acknowledgment info arrives)
send or multiply-send missing data
end;
end;
2nd Transmission Tactic
more efficient use of bandwidth resources (i.e., less retransmissions);
potentially produces more gaps in the receive window;
suitable for bulk data transfers.
send all application data
until (all application data is acknowledged)
exploit current acknowledgment info
send or multiply-send missing data
end;
DTTP Deployed in an IP-Enabled Internetwork
common network layer (IP in the figure) with potentially heterogeneous
underlying protocols
DTTP Deployed in a DTN-Enabled Internetwork
DTTP’s custody transfer functionality is deactivated (since offered by DTN);
DTTP is essentially used as a delay-tolerant, transport protocol
Simulation Parameters and Topologies
10MByte file transfer; last link
intermittent connectivity (70% on & 30% off)
Parameter
Value
{Round Trip Time (sec),
Data Rate (Mbps)}
{500, 10}
{1200, 2}
{2500, 0.5}
Number of Hops
2, 5
Packet Error Rate (%)
0, 1, 5, 10
2-hop and 5-hop Topologies
Simulation Results
File delivery completion time (using different communication times)
Simulation Results
RTT impact on file delivery completion time
Future Work
•
investigate various acknowledgment schemes (e.g. SACK, SNACK, other
mechanisms...) to mitigate bandwidth asymmetries
•
improve retransmission behavior (in relation to delay-bandwidth product;
incorporate relevant timers)
•
implement data forwarding via parallel paths, and explicit signaling for
storage resources exhaustion
•
explore network dynamics in space environments: e.g. buffer resources and
rate-based transmission trade-offs
...any questions?