Transcript Routing_DTN-tkkwon

Transportation-aware Routing in
Delay Tolerant Networks (DTNs)
Asia Future Internet 2008
Taekyoung Kwon
Seoul National University
outline
1 Introduction
2 Scenario Model
3 Our Approaches
4 Summary
2
Introduction
 DTN

Delay (or Disruption) Tolerant Networks
 Delay? Disruption?


Interplanetary networks
Sensor networks


Vehicular networks



Nodes sleep to save power
Mobile devices get out of other devices’ radio ranges
Opportunistic networks
 a sender and a receiver make contact at an unscheduled time
Underwater networks
Introduction
 Motivation

DTNs may have to be accommodated in future networks
 Intermittent connectivity
 Long or variable delay
 Asymmetric data rates
 Heterogeneous links
 High packet error rates
 Limited node uptime
Research Issues in DTNs
 Delay Tolerant Network Architecture

Overall redesign
 E.g. Bundle Protocol
 Routing Protocols


Delivery ratio
Reducing delay
 Congestion control
 Distributed Caching
 Multicast/Anycast
IP routing may not work
 E2e connectivity may not exist at the same time
 Routing (e.g. MANET) performs poorly in DTN
environments
 Some assumptions for routing will not work

E.g. BGP leverages TCP
Source: Kevin Fall, IRTF DTN RG
6
Related Work (mobility)
 Mobility model
DTN
No Mobility
Mobility
Routine
Predictable
Random
Tendencybased
Related work (routing)
 Some Routing Strategies


Epidemic routing
 Flooding
Spray and wait (S&W)
 Limited number of copies of a message
 Important Metrics



delivery probability
delivery latency
overhead ratio
Motivation
 Existing routing protocols use only past information like
contact history, etc.
 DTN Routing can leverage additional information in the
future

speed, direction, destination of mobile node, etc.
 We want to propose routing protocol using these
additional information
Scenario Model
 When to use DTN?

DTNs can be used for delay tolerant applications
 environmental monitoring, some publish/subscribe
applications
 We assume that each node has location information

E.g. GPS, Navigation, localization techniques
Potential Approaches
 Leveraging mobility information




Direction of mobile host
Speed of mobile host
Location of mobile host’s destination
Location of message’s destination
 Message’s destination can be fixed or mobile
 Our approaches



Direction-based
Destination-based
Transportation info-based
Our Approach 1
 Direction-Based routing protocol



Spray & Wait based
Number of tokens: n
Number of split tokens depends on direction difference
sender’s direction
0°
hand over
-n*angle/180° tokens
hand over
n*angle/180° tokens
hand over n/2 tokens
90 °
receiver’s direction
12
-90°
Our Approach 2
 Destination-Based routing protocol


Spray and wait based
Number of tokens for handover
 n/2*( distance / max diameter )
MAP
Receiver’s destination
Sender’s destination
13
Hybrid of approaches 1 and 2
 Direction-Distance-Hybrid (DDH)
Direction
Destination
Handed over tokens
similar
close
few
similar
far
medium
different
close
medium
different
far
n/2
 n/2*Direction(d1)*Distance(d2)*Speed(s)






Direction(): function ranged [0,1]
Distance(): function ranged [0,1]
Speed(): function ranged [0,1]
d1: direction difference of two nodes
d2: distance difference of two nodes’ destinations
s: difference of nodes’ speeds
14
Simulation results (1/2)
 Simulator


The Opportunistic Network Environment (ONE) simulator
http://www.netlab.tkk.fi/~jo/dtn/
 Parameter settings
Parameters
Value
Area size (m*m)
Number of nodes
4500 X 3400
100 (mobile), 10 (static)
Transmission range (m)
100
Speed (m/s)
0~18
Buffer size (GB)
1 (mobile), 200 (static)
Message size (MB)
0.01 ~ 3
Transmission rate (KB)
250
Movement model
Random waypoint
15
Simulation results (2/2)
 Comparison btw. S&W and DDH
 DDH can deliver 18% more packets than S&W
 When destination is fixed
* : # of delivered packets per 1000 relayed packets
16
Problem of Previous Approaches
 Randomization effect problem



It is caused by local view of tendency
As number of contacts is increased, direction or distance is
randomized
Effect of our proposal gets reduced
 An illustration


Some tokens can be carried in the
same direction
movement information that decides
the number of copies relayed
becomes meaningless
Angle = 90°
∴ handover n/2 tokens
1st contact
2nd contact
Angle = 90°
∴ handover n/4 tokens
Scenario Model
 A DTN area consists of a certain number of subareas or
regions
 There is a need of DTN between regions due to poor
infrastructure or delay tolerant application
 How to dissemination messages between regions
efficiently
Region 1
Region 2
Our Approach 3
 Prevention of randomization problem using history


Area is divided into several sub areas with non uniform distribution
Token handover policy
 When a source creates the message, it reserves a fixed number of
tokens for each sub-area
 If the source meets a mobile host toward other regions, it sends the
message to the host with pre-reserved tokens
 Tokens can be distributed more evenly across the area
19
Simulation Settings
 Simulator: Opportunistic Network Environment (ONE)
 Area size: 45 X 34km2

4 sub-areas (20x15km2 each)
 # of nodes: 500

Intra-area node & Inter-area node
 Tx range: 100m
 Speed: 100km/h, 4~60km/h
 S&W copies: 32
 Packet


# of packets: 1000 (2 packets per each node)
Packet size: ~ 30KB
 Buffer size big enough
Simulation Results
 Destination is mobile
 Delivery ratio
= # of delivered packets / # of originated packets
Delivery Probability
Delivery Probability (20% Inter-area Mobile Nodes)
0.6
0.5
0.4
epi_20
snw_20
our_20
0.3
0.2
0.1
0
0.5
1
1.5
Days
2
Simulation Results
 Overhead ratio
= (# of relayed - # of delivered) /
# of delivered
 Average number of relay
nodes
400
4.5
350
4
3.5
# of relayed nodes
Overhead Ratio
300
250
200
150
3
2.5
2
1.5
100
1
50
0.5
0
0
Epidemic
SprayAndWait
10%
20%
Region-based
Epidemic
SprayAndWait
10%
20%
Region-based
Simulation Results
 Avg. latency
 Med. latency
120000
115000
110000
115000
Latency Med.
Latency Avg.
105000
110000
105000
100000
95000
100000
90000
95000
85000
Epidemic
SprayAndW ait
10%
20%
Region-based
Epidemic
SprayAndW ait
10%
20%
Region-based
Conclusions
 DTNs may play a vital role in future
 Routing is a key player in DTNs
 We proposed



Direction-based
Distance-based
Transportation info-based
 Destination’s mobility affects the routing performance
 The more information, the better routing