On Using Probabilistic forwarding to Improve HEC

On Using Probabilistic Forwarding to
Improve HEC-based Data Forwarding in
Opportunistic Networks
Ling-Jyh Chen1, Cheng-Long Tseng2 and Cheng-Fu Chou2
1Academia Sinica
2National Taiwan University
Motivation
• There are numerous opportunistic networking
applications.
– wireless sensor network, underwater sensor
network, pocket switched network, people network,
and transportation network
• Traditional data forwarding algorithms are
not suitable for opportunistic networks.
– Scheduled optimal routing method
– Mobile relay approaches (Message ferry)
Related work
• Replication-based approaches
– The messages are replicated. Several
identical copies are transmitted over the
networks to mitigate the effects of a
single path failure.
– For example:
• Epidemic Routing,
• Controlled Flooding,
• mobility pattern-based scheme (Prophet)
Related work
• Coding-based approaches
– Transforming a message into another
format prior to transmission.
– For example:
• Erasure coding (EC), Aggressive Erasure Coding
(A-EC), Hybrid Erasure Coding (H-EC)
• Network Coding
Our Contribution
• We propose a message scheduling
algorithm, Probabilistic Forwarding, to
improve H-EC scheme.
• Using a set of simulations, we show the
proposed approach can provide better
data delivery performance.
Overview of H-EC
• Erasure Coding:
– Providing better fault-tolerance by adding
redundancy without the overhead of strict
replication.
• Reed-Solomon,
• Low-Density Parity-Check (LDPC) based coding
(Gallager, Tornado, and IRA codes)
Erasure Coding
A
A-1
B
A-3
A-2
A-4
B-1
C
B-3
B-2
(r,n)=(2,4)
B-4
C-1
D
C-3
C-2
C-4
D-1
D-3
D-2
Lossy Channel
A-1
A-3
A-2
A-4
A
B-1
B-3
C-1
D-1
B-2
B
C-4
C
D
D-4
Overview of H-EC
• H-EC: Hybrid of EC and A-EC
– First copy is sent using EC
– Second copy is sent using A-EC during the residual
contact duration after sending the first EC block
The Purposed Method: HEC-PF
• Probabilistic forwarding
– The HEC-PF scheme dost NOT enter the
aggressive forwarding phase unless a newly
encountered node has a higher likelihood
of successfully forwarding the message to
the destination node that the current
nodes.
• Delivery Probability
Delivery Probability
• Based on the observed contact history
• Take the contact frequency and contact
volume into consideration.
• The proportion of time that the two
nodes are in contact in the last T time
units.
the aggregated contact volume
Delivery
Probability
between the node pair Xi and Xj in
the last T time units
One-hop delivery probability
K: number of nodes in the network
The source
Node Xi: the i-th node
The ith
tXi;Xj:the aggregated
contact volumeNode
The Destination
between the node
Node pair Xi and Xj in the last
T time units
Delivery Probability
Two-hop delivery probability
Three-hop delivery probability
k-hop delivery probability
Probabilistic Forwarding
Evaluation
• DTNSIM: A Java-based DTN simulator
• Performance metric:
– Delay performance
– Transmission overhead
• Evaluating Scenarios:
Evaluation I: two-hop scenario
Evaluate the delay performance of the HEC-PF scheme for
message delivery.
Maximum message delivery distance (hops) H=2,
The transitive property of message delivery (hops) K=2
Power-Low Scenario
ZebraNet Scenario
UCSD Scenario
Evaluation II:
Variable k Scenarios
We evaluate the performance with various k values
(k = 2,3,4,5)
ZebraNet Scenario
UCSD Scenario
Evaluation II:
Variable k Scenarios
Evaluation III:
Variable H Scenarios
We evaluate the performance with various maximum
forwarding distance settings (H = 2,3,4,5)
ZibraNet Scenario
UCSD Scenario
Evaluation II:
Variable H Scenarios
Conclusion
• We purposes a new scheme for data forwarding by
incorporating the basic H-EC scheme with a new
feature, Probabilistic Forwarding.
• Using simulations as well as both synthetic and
realistic network traces, we show that the proposed
has better performance in terms of delivery latency
and completion ratio.
• We show that the completion ratio improves as the
maximum forwarding distance or the considered hop
distance of the delivery probability increases.
Thank You!