Chien-Shiu Lin, Wei-Shyh Chang, Ling-Jyh Chen, Ting-Kai Huang

Download Report

Transcript Chien-Shiu Lin, Wei-Shyh Chang, Ling-Jyh Chen, Ting-Kai Huang

Chien-Shiu Lin, Wei-Shyh Chang, Ling-Jyh Chen,
Cheng-Fu Chou, and Ting-Kai Huang
Motivation
 Delay /Disruption Tolerant Networks (DTNS)
attracted increasing attention recently.
 What’s the crucial factors for a successful
routing algorithms in Delay/Disruption Tolerant
Networks?
2
Our contributions

Analysis of several kinds of routing schemes




Epidemic routing
PROPHET routing
NC routing
Propose two enhancements for PROPHET


Contact utilization (PROPHET_CU)
Contact duration ratio (PROPHET_CD)
3
Epidemic routing
 “Store, carry and forward” model
 Utilize redundancy to increase the success
probability
 But,

flooding overhead
4
PROPHET routing
 Based on Epidemic Routing Scheme
 Use “delivery predictability” to reduce flooding
overhead
 Delivery predictability
 Pinit: an initialization constant
P( a ,b )  P( a ,b ) old  (1  P( a ,b ) old )  Pinit ,
Pinit  [0,1]
 Nodes that are often encountered have a high
delivery predictability.
5
PROPHET Enhancements (PROPHET_CU)
T
UP
Contact between A and C
DOWN
UP
Contact between B and C
DOWN
Time line
 The number of contacts vs. the duration of contact
 Contact Utilization (Prophet_CU)
PS1, D  TS , D / T
Total contact
duration of pair (S,D)
Fixed time interval
6
PROPHET Enhancements (PROPHET_CD)
T
A
B
C
D
C
C
Time line
 Super stars avoidance (load balance)
 Contact Duration Ratio( Prophet CD)
N
1
S ,D
P
 TS , D /  TS ,i
i 1
Total contact
duration of Node S
7
Network Coding
 Use coding skill to increase success probability
 Apply Network Coding on the packet
8
Simulation
 DTNSIM : A Java-based simulator
 A Real-world wireless trace
 ZebraNet1
 ZebraNet2 (decreasing 30% contact
durations)
ZebraNet1
ZebraNet2
Avg contact duration/pair/day(sec)
22,211
12,825
Avg # of contacts/pair/day
2.06
2.06
Number of nodes
34
34
9
Simulation Settings
Trace Name
ZebraNet1, ZebraNet2
CBR Traffic (1 MB/msg)
1 msg/hour
Contact bandwidth
1 MB/s
H (hop count limit)
5
m(computation depth)
4
T (time interval)
2 day
Number of pairs
1, 10, 20, 30
Buffer size
400 MB
Simulation time
16 day
Pinit / γ / ω
0.75 / 0.98 / 0.25
10
PROPHET vs. Enhanced PROPHET
100%
700000
95%
600000
90%
85%
PR
80%
75%
PR_cu
70%
PR_cd
transmission overhead
delivery ratio (%)
 ZabraNet1
65%
60%
500000
400000
PR
300000
PR_cu
200000
PR_cd
100000
0
1
10
20
30
1
number of node pairs
Node A
Node B
10
20
30
number of node pairs
CA,C
CA,D
CB,C
Time line
11
PROPHET vs. Enhanced PROPHET
1
PROPHET
PROPHET_CU
PROPHET_CD
Delivery Ratio CCDF
0.8
0.6
0.4
0.2
0
0
200000
400000
600000
30 node pairs
800000
1e+006
Latency (seconds)
12
PROPHET vs. Enhanced PROPHET
 ZebraNet2
300000
85%
delivery ratio
80%
75%
70%
PR
65%
PR_cu
60%
PR_cd
55%
50%
TRansmission Overhead
90%
250000
200000
PR
150000
PR_cu
100000
PR_cd
50000
0
1
10
20
number of node pairs
30
1
10
20
30
Num of Node Pairs
13
Routing comparison
1400000
90%
1200000
Delivery Ratio
80%
70%
Epidemic
60%
PR
50%
PR_cd
40%
NC(f=4)
30%
20%
1
10
20
Number of Node Pairs
30
Transmission Overhead
100%
1000000
800000
Epidemic
600000
PR
PR_cd
400000
NC(f=4)
200000
0
1
10
20
30
Number of Node Pairs
14
Routing comparison: various error rate
Delivery Ratio/
Delivery Ratio with No Error Rate
1.2
1
0.8
Epidemic
0.6
PR
PR_cd
0.4
NC(f=4)
0.2
0
0.01
0.05
0.1
0.15
0.2
Error Rate
16
Conclusion
 We propose enhancement schemes on PROPHET.
 Contact duration is an important factor
 We study several DTN routing schemes and
suggest suitable network environment for each
routing protocol.
 Epidemic routing is sensitive to network loading.
 NC has best delivery ration in congestion situation.
 NC routing is vulnerable in network with high error
rate
17
Thank you
18