© University of Alabama 1 Chapter 1: Identifying the Intertwined Links between Mobility and Routing in Opportunistic Networks Xiaoyan Hong Bo Gu University of Alabama ROUTING IN OPPORTUNISTIC NETWORKS
Feb 24, 2016
© University of Alabama 1
Chapter 1: Identifying the Intertwined Links between Mobility and Routing in
Opportunistic NetworksXiaoyan Hong
Bo Gu
University of Alabama
ROUTING IN OPPORTUNISTIC NETWORKS
© University of Alabama 2
OutlineIntroductionMobility modelsMobility characteristicsRouting protocolsFuture directionsSummary
© University of Alabama 3
MOTIVATION Mobility intertwines with routing protocols to play a vital
role in opportunistic networks
Mobility properties are utilized by routing protocols to improve performance
Study on mobility models, analytical results on motion characteristics and routing strategies will help developing novel integrated mobility and message dissemination solutions for opportunistic networks
© University of Alabama 4
INTRODUCTION Present a survey over mobility models, analytical
results on motion characteristics and routing strategies
Mobility models are the evaluation tools for routing protocols and the sources for movement pattern analysis
Analytical results contribute to new mobility models with increased flexibility in reproducing desired network scenarios
Routing protocols can make use of underlying mobile topological structures from results of mobility analysis
© University of Alabama 5
Intertwined Three Components
Spatial propertiesTemporal properties
Graph properties
Motion Characteristics
Proactive routingReactive routing -contact based -community based-auxiliary node based
Routing Protocols
no map, no intentionw map, no
intentionno map, w intention
Mobility Models
w map, w intention
New m
odel
Analys
is
Assist routing
Evaluation
© University of Alabama 6
Outline: Mobility Models
Spatial properties
Temporal propertiesGraph properties
Motion Characteristics
Proactive routingReactive routing -contact based -community based-auxiliary node based
Routing Protocols
no map, no intention
w map, no intentionno map, w intention
Mobility Models
w map, w intention
New model
Analysis
Assist routing
Evaluation
© University of Alabama 7
MOBILITY MODELS Movements are most likely the explicit or implicit
results of their social or personal activities. Physical locations Social intentions
Classifications Non-Map Without-Intention Models Map Without-Intention Models Non-Map With-Intention Models Map With-Intention Models
© University of Alabama 8
Non-Map Without-Intention Models Attributes:
no restrictions on paths nor intention of movement Basic model
Random Walk Model [8]: Memoryless Random Waypoint Model [28]: Delay factor to simulate pauses Random Direction Mobility Model[43]: Additionally deal with the
movements when hitting simulation boundary Realistic model
Gauss-Markov Mobility Model [30]: Simulate the acceleration and deceleration
Heterogeneous Random Walk[40]: Simulate the clustered network
© University of Alabama 9
Map Without-Intention Models Attributes:
movements are restricted to physical world paths.
Freeway model[1]: Vertical and horizontal tracks of freeway
City block[14]: Street grid Street Random Waypoint mobility model[11]:
Considering the intra-segment mobility and inter-segment mobility on street grid
Vehicular network model[44]: Stop signs, timed traffic lights and control on next road
© University of Alabama 10
Non-Map With-Intention Models Attributes:
No path restriction With individual or shared movement intentions
Group based model Reference Point Group Mobility Model (RPGM)[22]: paths of
nodes in the same group following the movement of the group leader
Interaction-based mobility model[34]: characterizes the formation and disaggregation of hot spots at random times and locations
Community based model Community based mobility model[35]: Captures the feature
that a number of hosts are grouped together Community model with cyclic pattern[54]: defines the repeating
time period to model re-visits to the same locations
© University of Alabama 11
Map With-Intention Models Attributes
Realistic features such as moving along paths and with intentions
Trace based model Bus traces[2], GPS trace[9], WLAN trace[51], Trace in campus
[23] Agenda Driven Mobility model[59]: use National Household
Travel Survey (NHTS) data to synthesize each node’s agenda, which derives its mobility of when, where and what (pause time)
Graph-based model Area Graph based mobility model[4]: A directed and weighted
graph to model locations and paths between locations Levy walk based model
Heavy-tail distribution[41]: movement increment is distributed according to a heavy-tail distribution
© University of Alabama 12
Summary of the Models Trend of mobility modeling has moved towards more
realistic by taking considerations of both social intentions and geographical features Artificially consider social interaction and attraction Analyzing real world traces WLAN associations give hits on mobility
Impact Effective evaluation tools Play an important role for message forwarding in opportunistic
networks
© University of Alabama 13
Outline: Mobility Characteristics
Spatial properties
Graph properties
Motion Characteristics
Proactive routingReactive routing -contact based -community based-auxiliary node based
Routing Protocols
no map, no intentionw map, no
intentionno map, w intention
Mobility Models
w map, w intention
New m
odel
Anal
ysis
Assist routing
Evaluation
Temporal properties
© University of Alabama 14
MOBILITY CHARACTERISTICSContribute to performance evaluation,
simulation calibration, routings protocol design
Classifications Characteristics of Flight Locality Distribution Temporal Characteristics Joint Spatial and Temporal Analysis Graph Characteristics
© University of Alabama 15
Characteristics of Flight Flight: the longest straight line trip from one location to
another
Flight length distribution can be heavy-tail, or exponential
Flight reflects the diffusivity of mobility
Models with different diffusivity Random Waypoint model, Brownian Motion, Levy Walk model
Impact Diffusive nodes are helpful for relaying messages to larger
areas
© University of Alabama 16
Locality Distribution Different movement patterns lead to various spatial
locality distributions
Distributions can be uniform or heterogeneous
Discussed models: Brownian-motion, Random Waypoint Model, Heterogeneous
Random Walk
Impact Cluster based routing is suitable in networks with
heterogeneous distribution
© University of Alabama 17
Temporal Characteristics Many properties have been analyzed:
encounter frequency, pause time, hitting time, meeting time, inter-contact time, filling time, scattering time
Impact Encounter history matters for choosing next forwarder Pause time, hitting time, meeting time, inter-contact time are
useful in estimating message delay and delivery rate Filling time and scattering time describe the dynamics of hot
spots, can be useful for cluster-based routing
© University of Alabama 18
Joint Spatial and Temporal Analysis Time and space are closely related in mobility
Trajectory similarity: Compute similarity using a set of metrics including Euclidean distance, etc.
Discussed models: Vehicular model[29], Mobyspace [27], location based time-dependent link analysis[20][21]
Impact Routing uses clusters or high similarity nodes Help to identify popular locations in mobile networks and
trajectory segments Calculate communication latency
© University of Alabama 19
Graph Characteristics Using graph properties to identify mobility patterns
Centrality [17] Degree centrality, closeness centrality, betweenness centrality
Social networks k-clique community, network connectivity
Discussed models: Clique community[25], Continuum framework [10]
Impact Node with higher centrality as forwarder, community helps to
group mobile nodes, connectivity analysis
© University of Alabama 20
Summary: Characteristic Analysis (I)Categories Mobility
CharacteristicsFeatures for Routing
Flight Length Longest straight line trip from one locationto next location; node diffusivity
Message forwarder adopts highdiffusive nodes for fast dissemination
Locality Distribution
Distribution of node positions during movingprocess is either uniform or heterogeneous
Cluster based routing is suitable in networkswith heterogeneous distribution
Temporal Characteristics
Encounter frequency, pause time, hittingtime, meeting time, inter-contact time, fillingtime, scattering time
Encounter history for choosing next forwarder; Estimatingmessage delay and delivery rate
© University of Alabama 21
Summary: Characteristic Analysis(II)Categories Mobility
CharacteristicsFeatures for Routing
Joint Spatial-Temporal
Time and location relationships of groups,trajectory similarity
Routing uses clusters or nodes with high similarity
Graph Characteristics
Degree centrality, closeness centrality, betweenness centrality,k-clique community
Nodes with higher centrality as forwarder;community helps to group mobile nodes;connectivity analysis and evolution for performance
© University of Alabama 22
Outline: Routing Protocols
Spatial propertiesTemporal properties
Graph properties
Motion Characteristics
Proactive routingReactive routing -contact based -community based-auxiliary node based
Routing Protocols
no map, no intentionw map, no
intentionno map, w intention
Mobility Models
w map, w intention
New
m
odel
Anal
ysis
Assist routing
Evaluation
© University of Alabama 23
ROUTING STRATEGIES Routing principle: store-carry-forward
Classifications:
Proactive Routing: with centralized or off-line knowledge about network
Reactive Routing: without a global or predetermined knowledge• Contact based routing: forward messages using the encounter
history• Community based routing: identify and rely on various clusters• Auxiliary node based routing: introduce mobile or static message
ferries
© University of Alabama 24
Proactive Routing knowledge such as contacts history, queuing length
and traffic demands Use a graph with time-varying delay and capacity
Discussed protocols: Framework of DTN routing which takes different levels of
network knowledge [26] Treat routing as a resource allocation problem[2] Link with contact probability calculated from cyclic movement
pattern [32] Routing assisted by static relay nodes deployed at critical
locations for cyclic movement pattern[19] Mobyspace with the assumption of full network knowledge [27]
© University of Alabama 25
Reactive: Contact based routing Forwarding decision is made when two nodes encounter
each other
Discussed protocols Epidemic routing [52]: forward to each contact PROPHET: employ a probabilistic metric called delivery
predictability [31] Spray and Wait protocol: broadcasts only a fixed number of
copies of message [49] Seek and Focus protocol: hybrid protocol which includes utility-
based routing and randomized routing [49]
© University of Alabama 26
Reactive: Community based routing Identify and use a special group of nodes
Better sociability Frequent contacts with the destinations Attached to a hot location
Discussed Protocols Distributed method to identify central nodes[13] Multicast routing [18] Island Hopping [46] Connected dominating set for VANET [33]
© University of Alabama 27
Reactive: Auxiliary node based routing
Introduce nodes specially designed for message relay, either mobile or static
Discussed routing Auxiliary node with Levy Walk pattern[47] Levy Walk searching[53] Mobile message ferry[57] Static throw box[58]
© University of Alabama 28
Summary: Routing Protocols Relationships among routing strategies, mobility
models and their characteristics
TABLE II and TABLE III summarize the following Categories Routing protocols Main routing strategies Mobility models and features Applicable environments
© University of Alabama 29
FUTURE DIRECTIONS Social network related analysis and its connection to
opportunistic networks
Movements within a real road system
Novel message dissemination schemes that explore new social network properties
Management of opportunistic networks, examples include extending coverage, capacity and traffic aggregation
© University of Alabama 30
CHAPTER SUMMARYThis chapter presents a survey over mobility
models, analytical results on motion characteristics, and routing strategies that largely rely on mobility in opportunistic networks
More important, it provides a systematical overview and identifies the intertwining connections among the three areas.
© University of Alabama 31
Mobility Characteristics
Movement Patterns
CHAPTER SUMMARY
Random walk,Random
waypoint,…
Manhattan Model,
Freeway model, …Group Mobility
model, community
based model,…
Trace based model, Graph
based model,…
Flight, locality, temporal characteristics,
joint spatial-temporal, graph features
Routing Schemes Proactive routing, reactive routing (contact based, community based, auxiliary node
based)))
Applications of Opportunistic Networks
Gossipmule, content spreading in mobile social networks,
opportunistic Internet access, rural area networks)
Abstraction
Mobility assistance Evaluation
Comm. support
© University of Alabama 32
Thanks for your attention!