Upload
buidiep
View
233
Download
3
Embed Size (px)
Citation preview
Chapter 4
Routing ProtocolsRouting Protocols
1
Overview
� Routing in WSNs is challenging due to the inherent
characteristics that distinguish these networks from other
wireless networks like mobile ad hoc networks or cellular
networks.
� First, due to the relatively large number of sensor nodes, it is not possible
to build a global addressing scheme for the deployment of a large number
of sensor nodes. Thus, traditional IP-based protocols may not be applied of sensor nodes. Thus, traditional IP-based protocols may not be applied
to WSNs. In WSNs, sometimes getting the data is more important than
knowing the IDs of which nodes sent the data.
� Second, in contrast to typical communication networks, almost all
applications of sensor networks require the flow of sensed data from
multiple sources to a particular BS.
� Third, sensor nodes are tightly constrained in terms of energy, processing,
and storage capacities. Thus, they require careful resource management.
2
Overview (cont.)
� Fourth, in most application scenarios, nodes in WSNs are generally
stationary after deployment except for, may be, a few mobile nodes.
� Fifth, sensor networks are application specific, i.e., design requirements
of a sensor network change with application.
� Sixth, position awareness of sensor nodes is important since data
collection is normally based on the location.
� Finally, data collected by many sensors in WSNs is typically based on � Finally, data collected by many sensors in WSNs is typically based on
common phenomena, hence there is a high probability that this data has
some redundancy.
3
Overview (cont.)
� These routing mechanisms have taken into consideration the
inherent features of WSNs along with the application and
architecture requirements. The task of finding and maintaining
routes in WSNs is nontrivial since energy restrictions and
sudden changes in node status (e.g., failure) cause frequent and
unpredictable topological changes.unpredictable topological changes.
� To minimize energy consumption, routing techniques proposed
for WSNs employ some well-known routing tactics as well as
tactics special to WSNs, e.g., data aggregation and in-network
processing, clustering, different node role assignment, and
data-centric methods were employed.
4
Outline
� 4.1 Routing Challenges and Design Issues in WSNs
� 4.2 Flat Routing
� 4.3 Hierarchical Routing
� 4.4 Location Based Routing
� 4.5 QoS Based Routing
� 4.6 Data Aggregation and Convergecast � 4.6 Data Aggregation and Convergecast
� 4.7 Data Centric Networking
� 4.8 ZigBee
� 4.9 Conclusions
5
Chapter 4.1
Routing Challenges and Design Routing Challenges and Design
Issues in WSNs
6
Overview
� The design of routing protocols in WSNs is influenced by
many challenging factors. These factors must be overcome
before efficient communication can be achieved in WSNs.
� Node deployment
� Energy considerations
� Data delivery model
Node/link heterogeneity� Node/link heterogeneity
� Fault tolerance
� Scalability
� Network dynamics
� Transmission media
� Connectivity
� Coverage
� Data aggregation/convergecast
� Quality of service7
Node Deployment
� Node deployment in WSNs is application dependent and
affects the performance of the routing protocol.
� The deployment can be either deterministic or randomized.
� In deterministic deployment, the sensors are manually placed
and data is routed through pre-determined paths.
� In random node deployment, the sensor nodes are scattered � In random node deployment, the sensor nodes are scattered
randomly creating an infrastructure in an ad hoc manner. If the
resultant distribution of nodes is not uniform, optimal
clustering becomes necessary to allow connectivity and enable
energy efficient network operation.
8
Energy Considerations
� Sensor nodes can use up their limited supply of energy
performing computations and transmitting information in a
wireless environment. Energy conserving forms of
communication and computation are essential.
� Sensor node lifetime shows a strong dependence on the battery
lifetime. In a multihop WSN, each node plays a dual role as lifetime. In a multihop WSN, each node plays a dual role as
data sender and data router. The malfunctioning of some sensor
nodes due to power failure can cause significant topological
changes and might require rerouting of packets and
reorganization of the network.
9
Data Delivery Model
� Time-driven (continuous)
� Suitable for applications that require periodic data monitoring
� Event-driven
� React immediately to sudden and drastic changes
� Query-driven
� Respond to a query generated by the BS or another node in the network
� Hybrid
� The routing protocol is highly influenced by the data reporting method in terms of energy consumption and route stability.
10
Node/Link Heterogeneity
� Depending on the application, a sensor node can have a
different role or capability.
� The existence of a heterogeneous set of sensors raises many
technical issues related to data routing.
� Even data reading and reporting can be generated from these
sensors at different rates, subject to diverse QoS constraints, sensors at different rates, subject to diverse QoS constraints,
and can follow multiple data reporting models.
11
Fault Tolerance
� Some sensor nodes may fail or be blocked due to lack of power,
physical damage, or environmental interference.
� It may require actively adjusting transmit powers and signaling
rates on the existing links to reduce energy consumption, or
rerouting packets through regions of the network where more
energy is available.energy is available.
12
Scalability
� The number of sensor nodes deployed in the sensing area may
be on the order of hundreds or thousands, or more.
� Any routing scheme must be able to work with this huge
number of sensor nodes.
� In addition, sensor network routing protocols should be
scalable enough to respond to events in the environment.scalable enough to respond to events in the environment.
13
Network Dynamics
� Routing messages from or to moving nodes is more
challenging since route and topology stability become
important issues.
� Moreover, the phenomenon can be mobile (e.g., a target
detection/ tracking application).
� On the other hand, sensing fixed events allows the network to � On the other hand, sensing fixed events allows the network to
work in a reactive mode while dynamic events in most
applications require periodic reporting to the BS.
14
Transmission Media
� The traditional problems associated with a wireless channel
may also affect the operation of the sensor network.
� In general, the required bandwidth of sensor data will be low,
on the order of 1-100 kb/s. Related to the transmission media is
the design of MAC.
� TDMA (time-division multiple access)� TDMA (time-division multiple access)
� CSMA (carrier sense multiple access)
15
Connectivity
� High node density in sensor networks precludes them from
being completely isolated from each other.
� However, may not prevent the network topology from being
variable and the network size from shrinking due to sensor
node failures.
� In addition, connectivity depends on the possibly random � In addition, connectivity depends on the possibly random
distribution of nodes.
16
Coverage
� In WSNs, each sensor node obtains a certain view of the
environment.
� A given sensor’s view of the environment is limited in both
range and accuracy.
� It can only cover a limited physical area of the environment.
17
Data Aggregation/Convergecast
� Since sensor nodes may generate significant redundant data,
similar packets from multiple nodes can be aggregated to
reduce the number of transmissions.
� Data aggregation is the combination of data from different
sources according to a certain aggregation function.
� Convergecasting is collecting information “upwards” from the � Convergecasting is collecting information “upwards” from the
spanning tree after a broadcast.
18
Quality of Service
� In many applications, conservation of energy, which is directly
related to network lifetime.
� As energy is depleted, the network may be required to reduce
the quality of results in order to reduce energy dissipation in
the nodes and hence lengthen the total network lifetime.
19
Routing Protocols in WSNs: A taxonomy
Network Structure Protocol Operation
Flat routing• SPIN
• Directed Diffusion (DD)
Negotiation based routing• SPIN
Multi-path network routing
Routing protocols in WSNs
20
• Directed Diffusion (DD)
Hierarchical routing• LEACH
• PEGASIS
• TTDD
Location based routing• GEAR
• GPSR
Multi-path network routing• DD
Query based routing• DD, Data centric routing
QoS based routing• TBP, SPEED, MERR
Coherent based routing• DD
Aggregation• Data Mules, CTCCAP
Reference
� J. N. Al-Karaki and A. E. Kamal, “Routing techniques in
wireless sensor networks: a survey,” IEEE Wireless
Communications, vol. 11, no. 6, pp. 6-28, Dec. 2004.
21
Chapter 4.2
Flat RoutingFlat Routing
22
Overview
� In flat network, each node typically plays the same role and
sensor nodes collaborate together to perform the sensing task.
� Due to the large number of such nodes, it is not feasible to
assign a global identifier to each node. This consideration has
led to data centric routing, where the BS sends queries to
certain regions and waits for data from the sensors located in certain regions and waits for data from the sensors located in
the selected regions. Since data is being requested through
queries, attribute-based naming is necessary to specify the
properties of data.
� Prior works on data centric routing, e.g., SPIN and directed
diffusion, were shown to save energy through data negotiation
and elimination of redundant.
23
4.2.1
SPINSPINSensor Protocols for Information via Negotiation
24
SPINMotivation
� Sensor Protocols for Information via Negotiation, SPIN
� a Negotiation-Based Protocols for Disseminating Information in Wireless
Sensor Networks.
� Dissemination is the process of distributing individual sensor
observations to the whole network, treating all sensors as sink
nodesnodes
� Replicate complete view of the environment
� Enhance fault tolerance
� Broadcast critical piece of information
25
SPIN (cont.) Motivation
� Flooding is the classic approach for dissemination
� Source node sends data to all neighbors
� Receiving node stores and sends data to all its neighbors
� Disseminate data quickly
� Deficiencies
� Implosion
� Overlap
� Resource blindness
26
SPIN (cont.) Implosion
A
CB
x x
Node
The direction
of data sending
The connect
between nodes
CB
D
x x
27
SPIN (cont.) Overlap
q
r
s
Node
(q,r) (s,r)
Node
The direction
of data sending
The connect
between nodesThe searching
range of the
node
A B
C
28
SPIN (cont.) Resource blindness
� In flooding, nodes do not modify their activities based on the
amount of energy available to them.
� A network of embedded sensors can be resource-aware and
adapt its communication and computation to the state of its
energy resource.energy resource.
29
SPIN (cont.) Sensor Protocols for Information via Negotiation
� Negotiation
� Before transmitting data, nodes negotiate with each other to overcome
implosion and overlap
� Only useful information will be transferred
� Observed data must be described by meta-data
Resource adaptation� Resource adaptation
� Each sensor node has resource manager
� Applications probe manager before transmitting or processing data
� Sensors may reduce certain activities when energy is low
30
SPIN (cont.) Meta-Data
� Completely describe the data
� Must be smaller than the actual data for SPIN to be beneficial
� If you need to distinguish pieces of data, their meta-data should differ
� Meta-Data is application specific
� Sensors may use their geographic location or unique node IDSensors may use their geographic location or unique node ID
� Camera sensor may use coordinate and orientation
31
SPIN (cont.) SPIN family
� Protocols of the SPIN family
� SPIN-PP
� It is designed for a point to point communication, i.e., hop-by-hop routing
� SPIN-EC
� It works similar to SPIN-PP, but, with an energy heuristic added to it
� SPIN-BC
� It is designed for broadcast channels
� SPIN-RL
� When a channel is lossy, a protocol called SPIN-RL is used where adjustments
are added to th SPIN-PP protocol to account for the lossy channel.
32
SPIN (cont.) Three-stage handshake protocol
� SPIN-PP: A three-stage handshake protocol for point-to-point
media
� ADV – data advertisement
� Node that has data to share can advertise this by transmitting an ADV with
meta-data attached
� REQ – request for data
� Node sends a request when it wishes to receive some actual data
� DATA – data message
� Contain actual sensor data with a meta-data header
� Usually much bigger than ADV or REQ messages
33
SPIN (cont.)
B
C
F
REQ
ADV
ADV
REQ
REQ
34
A
B
D
E
data
ADV
REQ
ADV
ADVREQ
REQ
REQ
data
datadatadatadata
SPIN (cont.) SPIN-EC (energy-conserve)
� Add simple energy-conservation heuristic to SPIN-PP
� SPIN-EC: SPIN-PP with a low-energy threshold
� Incorporate low-energy-threshold
� Works as SPIN-PP when energy level is high
� Reduce participation of node when approaching low-energy-
thresholdthreshold
� When node receives data, it only initiates protocol if it can participate in
all three stages with all neighbor nodes
� When node receives advertisement, it does not request the data
� Node still exhausts energy below threshold by receiving ADV
or REQ messages
35
SPIN (cont.) Conclusion
� SPIN protocols hold the promise of achieving high
performance at a low cost in terms of complexity, energy,
computation, and communication
36
SPIN (cont.) Reference
� J. Kulik, W.R. Heinzelman and H. Balakrishnan, “Negotiation-
based protocols for disseminating information in wireless
sensor networks,” Wireless Networks, Vol. 8, pp. 169-185, 2002.
37
4.2.2
Directed DiffusionDirected DiffusionA Scalable and Robust Communication Paradigm
for Sensor Networks
38
Directed Diffusion
� Properties of Sensor Networks
� Data-centric routing
� No central authority
� Resource constrained
� Nodes are tied to physical locations
� Nodes may not know the topology� Nodes may not know the topology
� Nodes may fail or move arbitrarily
39
Directed Diffusion (cont.)
� Directed Diffusion is an important milestone in the data centric routing research of sensor networks
� Data centric
� Individual nodes are unimportant
� Request driven
� The sinks requests data by broadcasting interests
� Sources satisfying the interest can be found
� Intermediate nodes route data toward sinks
� Localized repair and reinforcement
� Multi-path delivery for multiple sources, sinks, and queries
40
Directed Diffusion (cont.)
� Sinks broadcast interest to neighbors
� Initially specify a low data rate just to find sources for minimal energy
consumptions
� Interests are cached by neighbors
� Gradients are set up pointing back to where interests came
from from
� Once a source receives an interest, it routes measurements
along gradients
41
Directed Diffusion (cont.)
� Gradients from Source to Sink are initially small
� Increased during reinforcement
Source
Event Event
Source
Event
Sink
Interest propagation
Source Source
Sink
Initial gradients set up
Source
Sink
Data delivery along re-
inforced path
42
Interest Propagation
� Flood interest
� Constrained or Directional flooding based on location is possible
� Directional propagation based on previously cached data
Gradient
Source
Sink
Interest
Gradient
Event
43
Data Propagation
� Multipath routing
� Consider each gradient’s link quality
GradientEvent
Source
Sink
Interest
GradientEvent
44
Reinforcement
� Reinforce one of the neighbor after receiving initial data.
� Neighbor who consistently performs better than others
� Neighbor from whom most events received
GradientEvent
Source
Sink
Interest
GradientEvent
45
Negative Reinforcement
� Explicitly degrade the path by re-sending interest with lower data rate
� Time out: Without periodic reinforcement, a gradient will be torn down
GradientEvent
Source
Sink
Interest
GradientEvent
46
Design Considerations
� Design Space for Diffusion
47
Conclusions
� Directed Diffusion provides a data-centric communication
protocol between sink and sources.
� Directed Diffusion has some novel features - data-centric
dissemination, reinforcement-based adaptation to the
empirically best path, and in-network data aggregation and
caching.caching.
48
Reference
� C. Intanagonwiwat, R. Govindan, and D. Estrin, “Directed
Diffusion: A Scalable and Robust Communication Paradigm
for Sensor Networks,”in the Proceedings of the Sixth Annual
International Conference on Mobile Computing and Networks
(MobiCom’00), August 2000.
49
Chapter 4.3
Hierarchical RoutingHierarchical Routing
50
Overview
� In a hierarchical architecture, higher energy nodes can be used
to process and send the information while low energy nodes
can be used to perform the sensing in the proximity of the
target.
� Hierarchical routing is mainly two-layer routing where one
layer is used to select cluster heads and the other layer is used layer is used to select cluster heads and the other layer is used
for routing.
� Hierarchical routing (or cluster-based routing), e.g., LEACH,
PEGASIS, TTDD, is an efficient way to lower energy
consumption within a cluster and by performing data
aggregation and fusion in order to decrease the number of
transmitted messages to the base stations.
51
4.3.1
LEACHLEACHLow-Energy Adaptive Clustering Hierarchy
52
LEACH
� LEACH (Low-Energy Adaptive Clustering Hierarchy), a
clustering-based protocol that minimizes energy dissipation in
sensor networks.
� LEACH outperforms classical clustering algorithms by using
adaptive clusters and rotating cluster-heads, allowing the
energy requirements of the system to be distributed among all energy requirements of the system to be distributed among all
the sensors.
� LEACH is able to perform local computation in each cluster to
reduce the amount of data that must be transmitted to the base
station.
� LEACH uses a TDMA/CDMA MAC to reduce inter-cluster
and intra-cluster collisions.
53
LEACH (cont.)
� Sensors elect themselves to be local cluster-heads at any given
time with a certain probability. These cluster-head nodes
broadcast their status to the other sensors in the network.
� Each sensor node determines to which cluster it wants to
belong by choosing the cluster-head that requires the minimum
communication energy.communication energy.
� Once all the nodes are organized into clusters, each cluster-head creates a schedule for the nodes in its cluster.
� Being a cluster-head drains the battery of that node. In order to
spread this energy usage over multiple nodes, the cluster-head
nodes are not fixed; rather, this position is self-elected at
different time intervals.
54
LEACH Architecture
55
Dynamic Cluster
cluster-head nodes = C at time t1 cluster-head nodes = C’ at time t1 + d
All nodes marked with a given symbol belong to the same cluster, and
the cluster head nodes are marked with a
56
Algorithm Details
� Two main phases
� Set-up phase
� the clusters are organized and cluster heads are selected
� Steady-state phase
� the data transfers to the BS (Base Station)
57
Algorithm Details (cont.)
� Set-up phase
� Node n choosing a random number m between 0 and 1
� If m < T(n) for node n, the node becomes a cluster-head where
1 [ * mod(1 / )]( )
Pif n G
P r PT n
∈
−=
� where P = the desired percentage of cluster heads (e.g., P= 0.05), r=the
current round, and G is the set of nodes that have not been cluster-heads
in the last 1/P rounds. Using this threshold, each node will be a cluster-
head at some point within 1/P rounds. During round 0 (r=0), each node
has a probability P of becoming a cluster-head.
1 [ * mod(1 / )]( )
0 ,
P r PT n
otherw ise
−=
58
Algorithm Details (cont.)
� Set-up phase
� Cluster heads assign a TDMA schedule for their members
where each node is assigned a time slot when it can transmit.
� Each cluster communications using different CDMA codes to
reduce interference from nodes belonging to other clusters.
59
Algorithm Details (cont.)
� Steady-state phase
� All source nodes send their data to their cluster heads
� Cluster heads perform data aggregation/fusion through local transmission
� Cluster heads send them back to the BS using a single direct transmission
60
Conclusion
� Advantages
� Increases the lifetime of the network
� Even drain of energy
� Disadvantages
� Highly dynamic environments
� Nodes use single-hop communication� Nodes use single-hop communication
61
Reference
� W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy-efficient
communication protocol for wireless sensor networks”, Proceedings of the
33rd Hawaii International Conference on System Sciences, January 2000.
62
4.3.2
PEGASIS PEGASIS Power-Efficient Gathering in Sensor Information
Systems
63
PEGASIS
� Power-Efficient Gathering in Sensor Information Systems
(PEGASIS) is a near optimal chain-based protocol.
� In order to extend network lifetime, nodes need only communicate with
their closest neighbors and they take turns in communicating with the
base station.
� When the round of all nodes communicating with the base station ends, a
new round will start and so on.new round will start and so on.
� This reduces the power required to transmit data per round as the power
draining is spread uniformly over all nodes.
� Two main objectives for PEGASIS
� Increase the lifetime of each node by using collaborative techniques and
as a result the network lifetime will be increased.
� Allow only local coordination between nodes that are close together so
that the bandwidth consumed in communication is reduced.
64
Main Procedures
� PEGASIS assumes that each sensor node can be able to communicate with the BS directly. It also assumes that all nodes maintain a complete database about the location of all other nodes in the network.
� Greedy Algorithm to Construct Chain
� To construct the chain, start with the furthest node from the BS
� Add to chain closest neighbor to this node that has not been visited� Add to chain closest neighbor to this node that has not been visited
� Repeat until all nodes have been added to chain
� Node i (mod N) will take turns as the leader in round i, (N
represents the number of nodes), and then the leader transmits
data to BS.
65
Main Procedures (cont.)
C0
C3
C1
C2
BS
Chain construction using the greedy algorithm
66
Main Procedures (cont.)
� When a node dies, the chain is reconstructed in the same manner to bypass the dead node
� For gathering data in each round, each node receives data from one neighbor, fuses with its own data and transmits to the other neighbor on the chain
� In a given round, we can use a simple control token passing approach initiated by the leader to start the data transmission from the ends of the chain
67
Main Procedures (cont.)
token token
c2 is a leader
Token passing approach
68
Main Procedures (cont.)
� PEGASIS performs data fusion at every node except the end
nodes in the chain
� Each node will fuse its neighbor’s data with its own to generate
a single packet of the same length and then transmit to its other
neighborneighbor
69
Main Procedures (cont.)
token token
c2 is leaderOperational flow
70
PEGASIS Architecture
Base Station
End node
Leader
End node
71
Conclusion
� Advantages
� Minimizing the total sum of transmission distances
� Increase the lifetime of each node
� Disadvantages
� The single leader can cause higher delay
� Uneven drain of energy� Uneven drain of energy
� An extension to PEGASIS, called Hierarchical-PEGASIS was
introduced with the objective of decreasing the delay incurred
for packets during transmission to the BS.
72
References
� S. Lindsey and C. Raghavendra, “PEGASIS: Power-Efficient Gathering in
Sensor Information Systems,” IEEE Aerospace Conference, Vol. 3, pp. 3-
1125 to 3-1130, Big Sky, MT, USA, 9-16 Mar. 2002.
� S. Lindsey, C. Raghavendra, and K. Sivalingam, “Data gathering in sensor
networks using the energy*delay metric,” the IPDPS Workshop on Issues in
Wireless Networks and Mobile Computing, 2001.
73
4.3.3
TTDDTTDDTwo-Tier Data Dissemination
74
Basic Design
� Each sensor node is aware of its own location.
� Once a stimulus appears, the sensors surrounding it collectively
process the signal and one of them becomes the source to
generate data reports.
� Sinks (users) query the network to collect sensed data. It can be
multiple.multiple.
� In addition, TTDD design assumes that the sensor nodes are
aware of their missions.
� TTDD design uses a grid structure so that only sensors located
at grid points need to acquire the forwarding information.
75
Basic Design (cont.)
� The data source proactively builds a grid structure throughout
the sensor field and sets up the forwarding information at the
sensors closest to grid points.
� With this grid structure in place, a query from a sink traverses
two tiers to reach a source.
� The lower tier is within the local grid square of the sink's� The lower tier is within the local grid square of the sink's
current location (called cells), and the higher tier is made of the
dissemination nodes at grid points
� The sink floods its query within a cell, when the nearest
dissemination node for the requested data receives the query, it
forwards the query to its upstream dissemination node toward
the source.
76
Basic Design (cont.)
� Grid Lifetime
� A source includes a Grid Lifetime in the data announcement message
when sending it out to build the grid.
� If the lifetime elapses and the dissemination nodes on the grid do not
receive any further data announcements to update the lifetime, they clear
their states and the grid no longer exists.
77
Grid Construction
Source
Dissemination Node
78
Query and Data Forwarding
Source
Dissemination Node
Sink
Immediate Dissemination Node
79
Multiple Sinks
Source
Dissemination Node
Sink
Immediate Dissemination Node
Sink2
80
Multiple Sinks (cont.)
Source
Dissemination Node
Sink
Immediate Dissemination Node
Sink2
81
Trajectory Forwarding
Source
Dissemination Node
SinkPrimary agent (PA)
Immediate agent (IA)
82
Trajectory Forwarding (cont.)
Source
Dissemination Node
Sink
Immediate Dissemination Node
Primary agent (PA)
Primary agent (PA)
83
Grid Maintenance
Source
Dissemination Node
84
Grid Maintenance (cont.)
Source
Dissemination Node
85
Conclusion
� TTDD can enable efficient data dissemination in large-scale
wireless sensor networks with sink mobility.
� Instead of passively waiting for queries from sinks, TTDD
exploits the property of sensors being stationary and location-
aware to let each data source build and maintain a grid
structure in an efficient way.structure in an efficient way.
� Sources proactively propagates the existence information of
sensing data globally over the grid structure, so that each sink's
query flooding is confined within a local gird cell only.
� Queries are forwarded upstream to data sources along specific
grid branches, pulling sensing data downstream toward sink.
� TTDD is a good way to building an infrastructure in stationary
sensor networks.86
Reference
� F. Ye, H. Luo, J. Cheng, S. Lu, and L. Zhang, “A Two-
Tier Data Dissemination Model for Large-scale Wireless Sensor Networks,”
in Proceedings of the ACM/IEEE 6th International Conference on Mobile
Computing and Networking (MobiCom’02), 2002.
87
Chapter 4.4
Location Based RoutingLocation Based Routing
88
Overview
� Sensor nodes are addressed by means of their locations.
� The distance between neighboring nodes can be estimated on the basis of
incoming signal strengths.
� Relative coordinates of neighboring nodes can be obtained by
exchanging such information between neighbors.
� To save energy, some location based schemes demand that
nodes should go to sleep if there is no activity.
� More energy savings can be obtained by having as many
sleeping nodes in the network as possible.
� Hereby, two important location based routing protocols, GEAR
and GPSR, are introduced.
� Geographical and Energy Aware Routing (GEAR)
� Greedy Perimeter Stateless Routing (GPSR)
89
4.4.1
GEARGEARGeographical and Energy Aware Routing
90
Geographical and Energy Aware Routing (GEAR)
� The protocol, called Geographic and Energy Aware Routing
(GEAR), uses energy aware and geographically-informed
neighbor selection heuristics to route a packet towards the
destination region.
� The key idea is to restrict the number of interests in directed
diffusion by only considering a certain region rather than diffusion by only considering a certain region rather than
sending the interests to the whole network. By doing this,
GEAR can conserve more energy than directed diffusion.
� The basic concept comprises of two main parts
� Route packets towards a target region through geographical and energy
aware neighbor selection
� Disseminate the packet within the region
91
Energy Aware Neighbor Computation
� Each node N maintains state h(N, R) which is called learned cost to region R, where R is the target region
� Each node infrequently updates neighbor of its cost
� When a node wants to send a packet, it checks the learned cost to that region of all its neighbors
� If the learned cost of a neighbor to a region is not available, the � If the learned cost of a neighbor to a region is not available, the
estimated cost is computed as follows:
c(Ni, R) = αd(Ni, R) + (1-α)e(Ni)
where
α = tunable weight, from 0 to 1.
d(Ni, R) = normalized distance of neighbor to region
e(Ni) = normalized consumed energy at node i
92
Energy Aware Neighbor Computation (cont.)
� When a node wants to forward a packet to a destination, it
checks to see if it has any neighbor closer to destination than
itself
� In case of multiple choices it aims to minimize the learned cost
h(Ni, R)
� It then sets its own cost to:� It then sets its own cost to:
h(N, R) = h(Ni, R) + c(N, Ni)
c(N, Ni) = combination of remaining energy of N and Ni and the
distance between them
93
Forwarding Around Holes
F G H I J
K L T
C – T = 2B – T =
x5
5A B C D E
S
h(C,T) = h(B,T)+c(C,B)
94
α is set to 1. Initially, at time 0, at node S, among all neighbors of S, B, C, D
are closer to T than S. h(B,T)=c(B,T)= , h(C,T)=c(C,T)=2, h(D,T)=c(D,T)= .5 5
Recursive Geographic Forwarding
� Once the target region is reached, the packets are disseminated within the region by recursive geographic forwarding
� Forwarding stops when a node is the only one in a sub-region
95
Ni
Recursive Geographic Forwarding (cont.) Pathologies
� Inefficient Transmission
� Recursive geographic forwarding vs. Restricted flooding
ARecursive Geographic
Forwarding 3 times for sending
and 3 times for receiving =
Restricted flooding 1 times for
sending and 4 times for receiving
= consuming
5 units of energy
F
E B
C
D
and 3 times for receiving =
consuming 6 units of energy5 units of energy
96
Recursive Geographic Forwarding (cont.) Pathologies
� Non-Termination
� When network density is low compared to (sub) target region size
K
C
B
F
L
A
E H
97
Recursive Geographic Forwarding (cont.) Proposed solution for pathologies
� Node degree is used as a criteria to differentiate low density
networks from high density ones
� Choice of restricted flooding over recursive geographic
forwarding is made accordingly
98
Conclusion
� GEAR strategy attempts to balance energy consumption and
thereby increase network lifetime
� GEAR performs better in terms of connectivity after initial
partition
99
References� Y. Yu, D. Estrin, and R. Govindan, “Geographical and Energy-Aware Routing:
A Recursive Data Dissemination Protocol for Wireless Sensor Networks”, UCLA Computer Science Department Technical Report, UCLA-CSD TR-01-0023, May 2001.
� Nirupama Bulusu, John Heidemann, and Deborah Estrin. “Gps-less low cost outdoor localization for very small devices”. IEEE Personal Communications Magazine, 7(5):28–34, October 2000.
� L. Girod and D. Estrin. “Robust range estimation using acoustic and multimodal sensing”. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2001), Maui, Hawaii, October 2001.
� Nissanka B. Priyantha, Anit Chakraborty, and Hari Balakrishnan. “The cricket location-support system”. In Proc. ACM Mobicom, Boston, MA, 2000.
� Andreas Savvides, Chih-Chieh Han, and Mani B. Strivastava. “Dynamic fine-grained localization in adhoc networks of sensors”. In Proc. ACM Mobicom, 2001.
100
4.4.2
GPSRGPSRGreedy Perimeter Stateless Routing
101
Greedy Perimeter Stateless Routing (GPSR)
� Greedy Perimeter Stateless Routing (GPSR) proposes the
aggressive use of geography to achieve scalability
� GEAR was compared to a similar non-energy-aware routing
protocol GPSR, which is one of the earlier works in geographic
routing that uses planar graphs to solve the problem of holes
� In case of GPSR, the packets follow the perimeter of the planar � In case of GPSR, the packets follow the perimeter of the planar
graph to find their route.
� Although the GPSR approach reduces the number of states a
node should keep, it has been designed for general mobile ad
hoc networks and requires a location service to map locations
and node identifiers.
102
Algorithm & Example
� The algorithm consists of two methods:
greedy forwarding + perimeter forwarding
� Greedy forwarding, which is used wherever possible, and
perimeter forwarding, which is used in the regions greedy
forwarding cannot beforwarding cannot be
103
Greedy Forwarding (cont.)
� Under GPSR, packets are marked by their originator with their destinations’ locations
� As a result, a forwarding node can make a locally optimal, greedy choice in choosing a packet’s next hop
� Specifically, if a node knows its radio neighbors’ positions, the locally optimal choice of next hop is the neighbor geographically closest to the packet’s destinationgeographically closest to the packet’s destination
� Forwarding in this regime follows successively closer geographic hops, until the destination is reached
104
Greedy Forwarding (cont.)
D
x
y
105
Greedy Forwarding (cont.)
� A simple beaconing algorithm provides all nodes with their
neighbors’ positions: periodically, each node transmits a beacon
to the broadcast MAC address, containing only its own
identifier (e.g., IP address) and position
� Position is encoded as two four-byte floating point quantities,
for x and y coordinate valuesfor x and y coordinate values
� Algorithm jittered each beacon’s transmission by 50% of the interval B between beacons, such that the mean inter-beacon transmission interval is B, uniformly distributed in [0.5B, 1.5B]
� Upon not receiving a beacon from a neighbor for longer than timeout interval T, a GPSR router assumes that the neighbor has failed or gone out-of-range, and deletes the neighbor from its table
106
Greedy Forwarding (cont.) The Problem of Greedy Forwarding
D
v z
|xD|<|wD|and|yD|x will not choose to forward to w or y using greedy
x
w y
v z using greedy forwarding
void
xx
107
The Right-Hand Rule: Perimeters
� Mapping perimeters by sending packets on tours of them, using
the right-hand rule. The state accumulated in these packets is
cached by nodes, which recover from local maxima in greedy
forwarding by routing to a node on a cached perimeter closer to
the destination
� This approach requires a heuristic, the no-crossing heuristic, to � This approach requires a heuristic, the no-crossing heuristic, to
force the right-hand rule to find perimeters that enclose voids
in regions where edges of the graph cross
108
x
y
z
Planarized Graphs
� While the no-crossing heuristic empirically finds the vast
majority of routes in randomly generated networks, it is
unacceptable for a routing algorithm persistently to fail to find
a route to a reachable node in a static, unchanging network
topology
� Motivated by the insufficiency of the no-crossing heuristic, we � Motivated by the insufficiency of the no-crossing heuristic, we
present alternative methods for eliminating crossing links from
the network
109
Planarized Graphs (cont.) Relative Neighborhood Graph (RNG)
u vw
110
Planarized Graphs (cont.) Gabriel Graph (GG)
u v
w
111
Planarized Graphs (cont.)
Gabriel Graph (GG)
Relative Neighborhood Graph (RNG)
Original
112
Combining Greedy and Planar Perimeters
� All data packets are marked initially at their originators as greedy mode
� GPSR packet headers include a flag field indicating whether the packet is in greedy mode or perimeter mode
� Packet sources also include the geographic location of the destination in packets
� Only a packet’s source sets the location destination field, it is � Only a packet’s source sets the location destination field, it is left unchanged as the packet is forwarded through the network
� Upon receiving a greedy-mode packet for forwarding, a node
searches its neighbor table for the neighbor geographically
closest to the packet’s destination
� When no neighbor is closer, the node marks the packet into
perimeter mode
113
Combining Greedy and Planar Perimeters (cont.)
� GPSR packet header fields used in perimeter mode forwarding
114
Combining Greedy and Planar Perimeters (cont.)
D
Lp
Lf
e0
xIf forwarding node to D < Lp to D, returns a packet to greedy mode
115
Conclusion
� GPSR generates routing protocol traffic in a quantity independent of the length of the routes through the network
� GPSR generates a constant, low volume of routing protocol messages as mobility increases
� GPSR doesn’t suffer from decreased robustness in finding routes
116
References� B. Karp and H. T. Kung, “Greedy Perimeter Stateless Routing for
Wireless Networks”, Proc. 6th Annual ACM/IEEE Int'l. Conf. Mobile Comp. Net., Boston, MA, pp. 243-54, August 2000.
� G. G. Finn, “Routing and addressing problems in large metropolitan-scale internetworks”, Tech. Rep. ISI/RR-87-180, Information Sciences Institute, March 1987.
� S. Floyd and V. Jacoboson, “The synchronization of periodic routing messages”, IEEE/ACM Transactions on Networking, Vol. 2, pp. 122-136, April 1994.messages”, IEEE/ACM Transactions on Networking, Vol. 2, pp. 122-136, April 1994.
� B. Karp “Greedy perimeter state routing”, Invited Seminar at the USC/Information Sciences Institute, July 1998.
� J. Saltzer, D. P. Reed, and D. Clark, “End-to-end arguments in system design”, ACM Transactions on Computer Systems, Vol. 2, No. 4, Pages: 277-288, November 1984.
117
Chapter 4.5
QoS Based RoutingQoS Based Routing
118
Overview
� In QoS-based routing protocols, the network has to balance
between energy consumption and data quality.
� In particular, the network has to satisfy certain QoS metrics,
e.g., delay, energy, bandwidth, etc. when delivering data to the
BS.
119
Outline
� 4.5.1 TBP (QoS of Bandwidth)
� Ticket-Based Probing
� 4.5.2 SPEED (QoS of Transmission time)
� A Stateless Protocol for Real-Time Communication
� 4.5.3 MERR (QoS of Energy)4.5.3 MERR (QoS of Energy)
� Minimum Energy Relay Routing
120
4.5.1
TBP (Ticket-Based Probing)TBP (Ticket-Based Probing)
QoS of Bandwidth
121
Ticket-Based Probing
� There are numerous paths from source to destination, we shall
not randomly pick several paths to search
� We shall not use any flooding path-discovery approaches,
which may send routing messages to the entire network
� On the other hand, the flooding algorithms can handle
information imprecision but have prohibitively high overheadinformation imprecision but have prohibitively high overhead
� We want to make an intelligent hop-by-hop path selection to
guide the search along the best candidate paths
122
Ticket-Based Probing (cont.)
S
D
123
Ticket-Based Probing (cont.)
� A ticket is the permission to search one path. The source node issues a number of tickets based on the available state information
� More tickets are issued for the connections with tighter requirements
� Probes (routing messages) are sent from the source toward the destination to search for a low-cost path that satisfies the QoSrequirementdestination to search for a low-cost path that satisfies the QoSrequirement
� Each probe is required to carry at least one ticket
124
Ticket-Based Probing (cont.)
S
i
Dj
k
125
Ticket-Based Probing (cont.)
SD
A
C
3 3
3 2x
Demand = 3
B
C
E
3
32
2
2
6
5
x
126
Ticket-Based Probing (cont.)
SD
A
C
3 3
3 2
Demand = 4(1-1,3) (1-1,3)
B
C
E
3
22
2
2
6
5(1-2,1)
(1-2,1)
(1-2,1)
127
Ticket-Based Probing (cont.)
SD
A
C
3 3
3 2
(1.1,3) (1.1.1,2)
Demand = 4
B
C
E
3
22
2
2
6
5(1.2,1)
(1.1.2,1)(1.1.2,1)
(1.2,1)
(1.2,1)
128
Ticket-Based Probing (cont.)
T1
S D
129
T2
Ticket-Based Probing (cont.)
T2
T1
S D
130
Ticket-Based Probing (cont.)
xT2
T1
S D
x
131
Ticket-Based Probing (cont.)
SD
A
C
4 3
3 2
xDemand = 4
(1,4)
(2.1,3)
B
C
E
3
24
2
3
6
5x(2.2,1)
(2.1,3)
(2.1,3)
(2.1,3)
(2.2,1)
(2.2,1)
132
Conclusion
� The routing overhead is controlled by the number of tickets,
which allows the dynamic tradeoff between the overhead and
the routing performance. Issuing more tickets means searching
more paths, which results in a better chance of finding a
feasible path at the cost of higher overhead.
� A distributed routing process is used to avoid any centralized � A distributed routing process is used to avoid any centralized
path computation that could be very expensive for QoS routing
in large networks.
� This approach not only increases the chance of success but also
improves the ability to tolerate the information imprecision
because the intermediate nodes may gradually correct a wrong
decision made by the source.
133
Conclusion (cont.)
� Ticket-based probing scheme achieves a balance between the
single-path routing algorithms and the flooding algorithms. It
does multipath routing without flooding.
� The basic idea is to achieve a near-optimal performance with
modest overhead by using a limited number of tickets and
making intelligent hop-by- hop path selection.making intelligent hop-by- hop path selection.
134
References� S. Chen and K. Nahrstedt, “On finding multi-constrained paths,” in Proc.
IEEE ICC’98, pp. 874-879.
� R. Guerin and A. Orda, “QoS-based routing in networks with inaccurate information: Theory and algorithms,” in Proc. IEEE INFOCOM’97, Japan, pp. 75-83.
� Q. Ma and P. Steenkiste, “Quality-of-service routing with performance guarantees,” in Proc. 4th Int. IFIP Workshop Quality of Service, May 1997, pp. 115-126.1997, pp. 115-126.
� Z. Wang and J. Crowcroft, “QoS routing for supporting resource reservation,” IEEE J. Select. Areas Commun., Sept. 1996.
� S. Chen and K Nahrstedt, “Distributed Quality-of-Service Routing in Ad Hoc Networks,” IEEE J. Select. Areas Commun, vol.17, no. 8, pp. 1488-1505, Aug. 1999.
135
4.5.2
SPEED (QoS of Transmission time)SPEED (QoS of Transmission time)QoS of Transsion time
136
SPEEDMotivation
� Freshness of data
� Promptness of Command and Control
137
SPEED (cont.)Design Objectives
� Stateless Architecture
� Soft Real-Time
� Minimum MAC Layer Support
� QoS Routing and Congestion Management
� Traffic Load Balancing
� Localized Behavior
� Void Avoidance
138
SPEED (cont.)Architecture
139
� Neighbor Set of Node I
� NSi = {node| distance (node, node i ) ≤ R}
� Forwarding Set of Node I
� FSi (Destination) = {node ∈ NSi | L – L_next > 0}
SPEED (cont.)SNGF (Stateless Non-deterministic Geographic Forwarding)
L
j L-L_Next
NSFS
i D
m
k
140
SPEED (cont.) NFL (MAC Layer Feedback)
SELF NeighborsSELF NeighborsSELF NeighborsSELF Neighbors
� Delay Estimation: Delay= Round Trip Time – Receiver Side Processing Time
� On/Off Switch
� Back-Pressure Rerouting
Last Mile Process
SNGFBackpressure
ReroutingNFL
BeaconExchange
APIUniCast MultiCast AnyCast
MAC
DelayEstimation
Neighbor
Table
� Relay Ratio Control01 >∀−=
∑i
ieif
N
eKu
01 =∃= ieifu
- SNGFNeighbor
Nodes
BeaconBeaconBeaconBeaconMR Setpoint
Neighborhood Table
Delay Estimation Beacon
SELF NeighborsSELF NeighborsSELF NeighborsSELF Neighbors
MAC Feedback
Back Pressure Beacon
Relay RatioController
RelayRatiomiss
ratio
on/off
141
Backpressure Rerouting based on MAC Layer
Feedback & SNGF
7 11
SPEED
20
110
30
115
Delay
0.5s
0.1s
0.4s
0.1s
ID
9
7
10
3 Packet Destination
2
3
5
9
10
DelayBoo
Node 5's NTNode 5's NTNode 5's NTNode 5's NT
Packet
Source
Destination
142
Backpressure Rerouting based on MAC Layer
Feedback & SNGF
7
6
ID Delay
5 0.1S
7 0.4SNode 6's NTNode 6's NTNode 6's NTNode 6's NT Packet (to 4)
2
3
5
9
10
Delay BooID Delay
5 0.5S
2 0.1S
4 0.1SNode 3's NTNode 3's NTNode 3's NTNode 3's NT4
11
12Packet 1
Packet 1
Beacon
Packet 2
Packet 2
Packet 2
Packet 2
Packet 2
143
SPEED (cont.) Void Avoidance
� In a similar way, it deals with traffic congestion.
� Backpressure beacon (ID, Destination, Positive Infinity)
� Greedy: It may not find a path even if it exists in the worst case
Last Mile Process
SNGFBackpressure
ReroutingNFL
BeaconExchange
APIUniCast MultiCast AnyCast
MAC
DelayEstimation
Neighbor
Table
1
2
3
4 5
144
SPEED (cont.) Last Mile Process
� AreaMulticastSend(Center position, radius, deadline, packet)
� AreaAnyCastSend(Center position, radius, deadline, packet)
� UnicastSend(Global_ID,deadline,packet)
� SpeedReceive()
Last Mile Process
SNGFBackpressure
ReroutingNFL
BeaconExchange
APIUniCast MultiCast AnyCast
MAC
DelayEstimation
Neighbor
Table
� SpeedReceive()
145
Conclusion
� SPEED maintains a desired delivery speed across the network
through a novel combination of feedback control and non-
deterministic QoS-aware geographic forwarding
� This combination of MAC and network layer adaptation
improves the end-to-end delay and provides good response to
congestion and voidscongestion and voids
146
References
� T. Hea, J. A Stankovic, C. Lu, and T. Abdelzaher, “SPEED: a
stateless protocol for real-time communication in sensor
networks,” in Proc. IEEE International Conference on
Distributed Computing Systems, pp. 46-55, May 2003.
� G. S. Ahn, A. T. Campbell, A. Veres, and L.H. Sun. “SWAN:
Service Differentiation in Stateless Wireless Ad Hoc Networks,” Service Differentiation in Stateless Wireless Ad Hoc Networks,”
In Proc. IEEE INFOCOM'2002, June 2002.
147
4.5.3 MERR (Minimum Energy Relay Routing)MERR (Minimum Energy Relay Routing)
QoS of Energy
148
MERRSystem Model
� Since a railroad train has a global linear structure by nature, we
consider in this paper linear WSNs as sensor networks having,
roughly, a linear topology
� such as sensors embedded in the outer surface of a pipeline or mounted along the
supporting structure of a bridge
� Aiming at such networks, we introduce two routing schemes
that efficiently utilize energy: Minimum Energy Relay Routing
(MERR) and Adaptive MERR (AMERR)
149
MERR (cont.)Energy Model
)(),(
)(
)(),(
γ
γ
drdrP
rrP
drdrP
txrxrelay
rxrx
txtx
εαα
α
εα
++=
=
+=
)(),( γεαα drdrP txrxrelay ++=
)( γεα dr +≡
150
)(),( drdrP txrxrelay εαα ++=
)( γdr εα +≡
γ
γ )1( −=
=
ε
αchar
charchar
opt
d
d
Dor
d
DK Kopt: optimal number of hops
MERR (cont.)Energy Model
� The base station is assumed to have unlimited energy supply.
� For a sensor to transmit a bit-stream of rate r over a distance d,
the transmitter power Ptx(r, d) is
� where αtx is the energy per bit consumed in the transmitter circuit, and ε
)(),( γεα drdrP txtx +=
� where αtx is the energy per bit consumed in the transmitter circuit, and ε
accounts for the energy dissipated in the transmit amplifier. The path loss
exponent γ typically ranges between 2 and 6; it is closer to 2 if there is a
perfect line-of-sight between transmitter and receiver and can go up to 6
in dense urban areas.
� The power Prx(r) needed to receive a bit-stream of rate r is
� where αrx is the energy per bit consumed by the receiver circuit.
151
txrx rrP α=)(
MERR (cont.)Energy Model
� For a sensor to receive a bit-stream of rate r and to forward it a
distance d onward, the power consumption is given
)(),( γεαα drdrP txrxrelay ++=
)( γεα dr +≡
� As Ptx, Prx, and Prelay scale linearly with r, we omit this term in
the following and implicitly assume r=1 bit/s.
152
MERR (cont.)Minimum Energy Path
� Suppose that a sensor S is located at distance D from the base
station BS and that S wants to deliver some data to BS. The
goal is to minimize the power needed on the entire path from S
to BS.
153
MERR (cont.)Minimum Energy Path
� S should transmit directly to BS if
� Otherwise, it is best to select (Kopt − 1) equally spaced,
intermediate nodes for retransmission. Kopt is the optimal
number of hops which is
γγεα /11 )))21(/(( −−≤D
number of hops which is
� where dchar is the characteristic distance
154
γ
γ )1( −=
=
ε
αchar
charchar
opt
d
d
Dor
d
DK
MERR (cont.)System Model
� We consider a linear WSN to be a sensor network having,
roughly, a linear topology.
� Data propagation is assumed to be unidirectional.
155
MERR (cont.) Minimum Energy Relay Routing
|| chardDE − || chardCD −= || chardBD −=‘
BSA B C D E
156
AMERRAdaptive Minimum Energy Relay Routing
157
Conclusion
� Linear topology networks often appear in pipeline monitoring and structural health monitoring, but the application that motivated this work is telemetry and control for freight railroad trains
� Using sensor networks to provide more timely information, the goal of this commercial application is to attain greater visibility of the rolling assets and cargo to allow for real-time failure of the rolling assets and cargo to allow for real-time failure prediction of a train’s components
158
References
� M. Zimmerling, W. Dargie, J. Reason, “Energy-Efficient Routing in Linear
Wireless Sensor Networks,” The Fourth IEEE International Conference on
Mobile Ad-hoc and Sensor Systems, Pisa, Italy, 8-11 October, 2007.
� M. Zimmerling, W. Dargie, and J. Reason, “Localized power-aware routing
in linear wireless sensor networks,” In CASEMANS ’08: Proceedings of the
2nd ACM international conference on Context-awareness for self-managing
systems, pp. 24-33, New York, 2008,.systems, pp. 24-33, New York, 2008,.
159
Chapter 4.6
Data Aggregation and ConvergecastData Aggregation and Convergecast
160
Outline
� 4.6.1 The Impact of Data Aggregation
� 4.6.2 Data Mules
� 4.6.3 Convergecasting Tree Construction and Channel
Allocation Problem (CTCCAP)
� 4.6.4 Distributed Time-Optimal Scheduling for Convergecast
161
Outline
� 4.6.1 The Impact of Data Aggregation
� 4.6.2 Data Mules
� 4.6.3 Convergecasting Tree Construction and Channel
Allocation Problem (CTCCAP)
� 4.6.4 Distributed Time-Optimal Scheduling for Convergecast
162
4.6.1 The Impact of Data Aggregation
� Impact of Data Aggregation in Wireless Sensor Networks
(Krishnamachari, Estrin, & Wicker, 2002)
� Aggregation in Sensor Networks
� Theoretical Results on Aggregation
� Aggregation Techniques
� Performance study� Performance study
163
Aggregation in Sensor Networks
� Traditional Address-Centric routing
� IP address routing
� Not suitable in large scale sensor networks
� Data-Centric Routing
� Content-based routing
� Enhance the data aggregation opportunity� Enhance the data aggregation opportunity
164
Source 1 Source 2
A B
Sink
Source 1 Source 2
A B
Sink
a) Address-Centric (AC) Routing b) Data-Centric (DC) Routing
1
1
2
2
21
1+2
DataAggregation
Theoretical Results on Aggregation
� Let there be k sources located within a diameter X, each a distance di from
the sink. Let NA and ND be the number of transmissions required with AC
and optimal DC protocols, respectively.
1. The following are bounds on ND:
( 1) min( )
min( ) ( 1)
D i
D i
N k X d
N d k
≤ − +
≥ + −
2. Asymptotically, for fixed k, X, as d = min(di) is increased,
3. Although the problem is NP-hard in general, the optimal data aggregation
tree can be formed in polynomial time when the sources induce a
connected subgraph on the communication graph.
min( ) ( 1)D i
N d k≥ + −
1lim D
d
A
N
N k→∞ =
165
Aggregation Techniques
� In general the formation of the optimal aggregation tree is NP-
hard. Some suboptimal DC routing heuristics as follows:
� Center at Nearest Source (CNSDC)
� All sources send the information first to the source nearest to the sink, which
acts as the aggregator.
� Shortest Path Tree (SPTDC)
Opportunistically merge the shortest paths from each source wherever they � Opportunistically merge the shortest paths from each source wherever they
overlap.
� Greedy Incremental Tree (GITDC)
� Start with path from sink to nearest source. Successively add next nearest
source to the existing tree.
� Address Centric (AC)
� No aggregation, distinct shortest paths from each source to sink.
166
Performance Study
Event-Radius model Random Sources model
167
Performance Study (cont.)
Energy Costs
Event-Radius model Random Sources model
168
Conclusions
� Data aggregation can result in significant energy savings for a
wide range of operational scenarios.
� The gains from aggregation are paid for with potentially higher
delay. It should be possible to design routing algorithms for
sensor networks in which this tradeoff is made explicitly.
169
Reference
� Bhaskar Krishnamachari, Deborah Estrin, and Stephen B. Wicker, "The
impact of data aggregation in wireless sensor networks," In Proceedings of
the 22nd International Conference on Distributed Computing Systems
Workshops (ICDCSW'02), pp. 575-578, Vienna, Austria, July 02-05 2002.
170
Outline
� 4.6.1 The Impact of Data Aggregation
� 4.6.2 Data Mules
� 4.6.3 Convergecasting Tree Construction and Channel
Allocation Problem (CTCCAP)
� 4.6.4 Distributed Time-Optimal Scheduling for Convergecast
171
4.6.2 Data Mules
� Data MULEs: Modeling a Three-tier Architecture for Sparse
Sensor Networks (Shah, Roy, Jain, & Brunette, 2003)
� Three-tier architecture
� Data MULEs approaches
172
Three-tier Architecture
Access points – ample resources
Mobile nodes – renewable resources for
intermittent connectivity
Source nodes – limited resources
173
� A top tier of WAN connected devices
� A middle tier of mobile transport agents
� A bottom tier made of fixed wireless sensor nodes
Data MULEs approach
� Data Mules
� Exploit mobile nodes (called MULEs)
� MULEs collect data when near sensor
� Transfer data to an access point when close
174
access pointsensor
Discussions
� Benefits
� Energy efficient
� Short distance communication
between sensor and MULE
� Scalable
� Addition of sensors or MULEs
requires no configuration
� Limitations
� No guarantees on data delivery
� MULEs may lose data
� MULEs may not arrive at a
sensor
� MULEs may not arrive at an
access pointrequires no configuration
� Simple
� Least functionality in sensors
� No forwarding, no global
discovery
access point
175
Reference
� Rahul C. Shah, Sumit Roy, Sushant Jain, and Waylon Brunette, "Data
MULEs: modeling and analysis of a three-tier architecture for sparse sensor
networks," Ad Hoc Networks, Volume 1, Issues 2-3, pp. 215-233, September
2003.
176
Outline
� 4.6.1 The Impact of Data Aggregation
� 4.6.2 Data Mules
� 4.6.3 Convergecasting Tree Construction and Channel
Allocation Problem (CTCCAP)
� 4.6.4 Distributed Time-Optimal Scheduling for Convergecast
177
Convergecasting in WSN
� WSN are mainly used for monitoring
� Monitoring involves data collection and request dissemination
� Convergecasting� Process of data collection from all or a set of sensors in the network
towards the base station (Many to one communication)
� Energy and latency minimization is required for WSNs
178
Convergecasting
� Route construction plays a major role during convergecasting
� Criterion for route construction
� Energy consumption
� Latency incurred
� Choice of MAC layer – since traffic is many to one
179179
Collisions
� Results in packet loss
� Need reliability, use retransmissions
� Retransmission increases energy consumption and latency
� Avoided by using a contention based or contention free MAC
protocolBS
180
1
BS
2
3 4 5
6
CollisionCollision CollisionCollision Coverage Area
or
Sensing Range
Data
Data
Energy & LatencyEnergy
� Energy consumed at a node is used for� Running the transceiver circuitry for transmitting a bit (Etrx)
� Amplifying a bit of data to be transmitted (Eamp)
It depends on the transmission distance
BS BS
E = 4nj Eamp= 4nj
181
3 hops
2 hopsP1
P2
P1
P2
n n
Eamp= 4nj
Eamp= 4nj
Eamp= 5nj
Eamp= 4nj
Eamp= 5nj
Energy consumed for running transceiver
To transmit k data bits from n to BS
P1: 3 * Etrx * k
P2: 2 * Etrx * k
Amplification energy consumed for
transmitting k data bits from n to BS
P1: (4nj + 4nj + 5nj) * k = 13 * k nj
P2: (4nj + 5nj ) * k = 9 * k nj
P1 and P2: paths
BS: Base Station
Energy & Latency (cont.) Energy
� Transceiver startup time
� Frequently switching the transceiver on leads to higher energy wastage
� Aggregation reduces packet header overhead
Time-slot =1 Time-slot = 2
Time-slot = 3
BSBS – Base Station
Aggregation reduces
transmitter startup energy
wastage
Slots allocated to children should
reduce cumulative startup time
of parent’s receiver 1
4 5
2
3
182
Energy & Latency (cont.) Latency
� Time taken to gather data at the base station
� Latency = No. of time-slots × Length of one-slot
� Balanced tree helps in reducing total number of time-slots and length of time-slots
Unbalanced Tree Balanced Tree
BS : Base StationBS
183
Number of slots = 4
Length of each slot = 4 packets
Latency = 16 units
Number of slots = 3
Length of each slot = 3 packets
Latency = 9 units
BS : Base Station
t =1 t = 2 t = 3
t = 4t = 1
BS
1
4 5
2
3
t =1 t = 2 t =1
t = 2t = 3
BS
1
4 5
2
3
Energy & Latency (cont.) Summary
� Energy and latency minimized by avoiding collisions
� Energy consumption can also be minimized by
� Reducing the number of hops
� Choosing path that minimizes amplification energy
� Reducing energy wastage due to transceiver startup time by performing
data aggregationdata aggregation
� Latency minimization by building a balanced routing tree
184
Main Procedures
� Assumptions
� Etrx < Eamp
� One transceiver per node
� Nodes have maximum transmission range (MEamp)
� Clock synchronization mechanism exists
� Builds the tree and allocates channel for the nodes� Builds the tree and allocates channel for the nodes
� Allocates channel for two different convergecast patterns
� Synchronous: Used for realtime data. Enables aggregation. Therefore
parent transmits after it receives from children (parent time-slot > child
time-slots)
� Asynchronous: Used for non-realtime data. Enables aggregation only if
data does not depend in time.
185
Synchronous Convergecast
� Data collection starts from leaf nodes
� Each parent waits for data from its children before sending its data
� Reordering based on timestamp is not necessary at base station
2 3
BS BS
<4, 1> <3, 1>
Note: Weights indicate the
amplification energy expended
Network
13
3
3
1.4
2.2
2
1 2
3
4
5
Convergecast Tree
1 2
3
4
5
<4, 1> <3, 1>
<3, 1>
<2, 1>
<1, 1>
amplification energy expended
to transmit a data bit over that
link
186 <t, c>: a tuple of time-slot t and CDMA code c
Asynchronous Convergecast
� Data collection takes place at independent and not interfering parts of the
network
� Reordering necessary at base station
� Latency will be low
Network 2 3
BS
Convergecast Tree
BS
Note: Weights indicate the Network
1
2 3
3
3
3
1.4
2.2
2
1 2
3
4
5
1 2
3
4
5
<1, 1>
<2, 1>
<3, 1>
<1, 1><2, 1>
Note: Weights indicate the
amplification energy expended
to transmit a data bit over that
link
187
Channel Allocation Criterion 1
� Each node has one transceiver
� Therefore a parent with two children cannot receive from both
of them at the same time using two different codes
� Therefore children transmit at different time instants
ParentX = ParentYParentX = ParentY
X Y
ParentX = ParentY
X Y
ParentX = ParentY
<t1,c1> <t2,c1><t1,c1> <t1,c2>
188
Channel Allocation Criterion 2
� Avoid exposed terminal problem
ParentX ParentYTransmission
range
If X and Y useCollisionCollision CollisionCollision
X Y
If X and Y use
the same channel
If X and Y transmit
at different time-slots
If X and Y transmit
using different CDMA
codes
189
Channel Allocation Criterion 3
� Parent cannot receive the same time it is transmitting
� Therefore, we have parent time-slot ≠ child time-slot
ParentParentxParentParentx
X
ParentX
<t1,c1>
<t1,c2>
X
ParentX
<t1,c1>
<t2,c1>
190
Algorithm
� Build the tree and allocate the channel (is a tuple of time-slot tand CDMA code c, <t, c>)
� Tree constructed in a top down manner
� Use channel allocation criteria defined earlier
� Additional criterion for synchronous convergecasting� child time-slot < parent time-slot
� Since tree construction is top down, it is not possible to allocate � Since tree construction is top down, it is not possible to allocate a valid time-slot for children
� Channel allocation in two phases� Phase I
� Construct tree and allocate channel in increasing order of time-slots
� Phase II� Reverse mapping of time slots to enable synchronous convergecast
191
CTCCAA: Phase I (Tree Construction)
� Construct the tree by reducing number of hops and then choose the path that consumes minimum amplification energy� Reason: Etrx < Eamp since transmission range of sensors are small
� Start constructing the tree with Base station (BS) as the root node
� Maintain a possible parent and a possible child list� Possible Parent List (PPL) = {All nodes recently added to the tree} � Possible Parent List (PPL) = {All nodes recently added to the tree}
� Possible Children List (PCL) = {x | there exists y ∈ PPL such that Eamp(x,y) < MEamp}
� Parent selection� For all x ∈ PCL parentx = arg Minforall y ∈ PPL Eamp(x,y)
� If for all x ∈ PCL parentx ≠ null, then copy PCL to PPL
192
Example: Phase I
Initially current level is 0
PPL = {BS}
PCL = {1, 2}
Since BS is the only possible
parent both 1 and 2 choose 1
2 3
BS
1 2
Weights on links indicate the
amplification energy expended
to transmit a data bit
parent both 1 and 2 choose
BS as their parent.
13
3
3
1.4
2.2
23
4
5
193
CTCCAA: Phase I (Channel Allocation)
� Use a combination of CDMA codes and time-slots
� Allocate children a time-slot that is greater than parent (will do
reverse mapping in phase II)
194
Example: Phase I
Weights on links indicate the
amplification energy expended
to transmit a data bitThis example assumes channel to be
divided over time.
Initially current level is 0
PPL = {BS}
BS
1<1, 1>PPL = {BS}
PCL = {1, 2}1
2
3
4
5
<1, 1><2, 1>
13
3
3
1.4
2.2
2
195
Example: Phase I
Weights indicate the
amplification energy
BS
1
Initially current level is 0
PPL = {1, 2}
PCL = {3, 4, 5}
<1, 1> 12
3
4
5
<1, 1>
<3, 1>
<4, 1>
<2, 1>
<2, 1>
13
3
3
1.4
2.2
2
196
CTCCAA: Phase II
� Only executed for synchronous convergecast
� Use maximum time-slot (Maxts) allocated in the network
� Actual time-slot = Maxts – allocated time-slot
BSMaxts = 4
197
<1, 1> 2
3
4
5
<2, 1>
<4, 1>
<3, 1>
<2, 1> <1, 1>
<4, 1>
<3, 1>
<2, 1>
<3, 1>1
Example
� This shows the advantage of divided channel over time and
CDMA Codes. CDMA codes help in reducing latency by
increasing time-slot reuse
BS
1 2
3
4
5
<3, 2><2, 2>
<1, 1>
<2, 1>
<1, 2>
198
Conclusions
� Convergecast will be preceded by broadcast in monitoring applications
� Measured energy and latency incurred during convergecastover a broadcast tree and a tree constructed by CTCCAA
� Similarly we measured energy and latency for broadcasting over both the trees
199
Reference
� V. Annamalai, S.K.S. Gupta, and L. Schwiebert, “On tree-based
convergecasting in wireless sensor networks,” IEEE Wireless
Communications and Networking, vol. 3, pp.1942-1947 , March 2003.
200
Outline
� 4.6.1 The Impact of Data Aggregation
� 4.6.2 Data Mules
� 4.6.3 Convergecasting Tree Construction and Channel
Allocation Problem (CTCCAP)
� 4.6.4 Distributed Time-Optimal Scheduling for Convergecast
201
Outline
� 4.6.4. Distributed time-optimal scheduling for convergecast
� 4.6.4.1. System model and Assumptions
� 4.6.4.2. Convergecast in tree networks
� 4.6.4.2.1. Linear Networks
� 4.6.4.2.2. Multi-line Networks
� 4.6.4.2.3. Tree Networks
� 4.6.4.2.4. Sleep schedule for energy conservation
� 4.6.4.3. Convergecast in general networks
� 4.6.4.4. Convergecast in other scenarios
202
4.6.4. Distributed Time-optimal Scheduling for
Convergecast
� Convergecast is a typical many-to-one communication pattern
in sensor network applications.
� In convergecast many, or all nodes in the network send data to
a base station during a relatively short time period.
� Using CSMA MAC layer, the convergecast latency incurred by
radial coordination is far from the optimal.radial coordination is far from the optimal.
� TDMA schedule such that the entire convergecast can be
completed in minimal number of timeslots.
203
4.6.4.1. System Model and Assumptions
� Assumptions
� the nodes and the associated base station are static
� the nodes (including the base station) cannot transmit and receive at the
same time
� the bandwidth of every wireless link in the network is assumed to be the
same
the network connectivity is fixed over time� the network connectivity is fixed over time
� the maximum length of a packet is fixed
� the drift in the clock of a node is bounded all the time.
204
4.6.4.2. Convergecast in Tree Networks
� There are three considers of sensor nodes and propose an
optimal convergecast scheduling algorithm.
� linear networks
� multi-line networks
� tree networks
205
4.6.4.2.1. Linear Networks
� We define the following states that a node can be in each
timeslot during the convergecast
� R: The node may receive from a neighboring node.
� T: The node can transmit.
� I : The node neither transmits nor receives.
206
Linear Networks (cont.)
A Linear Network
Convergecast Schedule
207
4.6.4.2.2. Multi-line Networks
� Convergecast scheduling algorithm for multi-line networks is
to schedule transmissions parallelly along multiple branches
Next timeslot
R T
208
State transition for convergecast scheduling
Next timeslotNext timeslot
R T
I
� The network consists of branches A, B, C and D (A < B < C <
D) with 3, 2, 2, and1 nodes
Multi-line Networks (cont.)
A multi-line network
209
Multi-line Networks (cont.)
Convergecast schedule for multi-line networks
210
The Pkts Left field is used to track the number of packets remaining in each branch.
The Last Slot field shows the last timeslot in which a branch has forwarded a
packet to the base station (two less than the last active timeslots).
4.6.4.2.3. Tree Networks
� Convergecast scheduling algorithm for tree networks is based
on the observation that a tree network can be reduced to a
multi-line network with each line represented as a combination
of linear branches of nodes.
211
Tree Networks (cont.)
Reduction of a tree network into linear branches.
(a) (b)
212
4.6.4.2.4. Sleep Schedule for Energy
Conservation
� Energy spent in sleep state is negligible.
� At most 3N timeslots are required to finish the convergecast.
� The total energy consumption in the network is
Joules. Hence, we conclude that the sleep schedule
results in about 50% energy conservation in linear networks.
2
)1(3 eNN +
213
4.6.4.3. Convergecast in General Networks
214
4.6.4.4. Convergecast in Other Scenarios
� We show that our convergecast scheduling algorithm is
applicable even when these assumptions do not hold.
� Base station initiated convergecast
� Nodes with multiple packets
� Non-ideal radio characteristics
215
Convergecast in Other Scenarios (cont.)
Base station initiated convergecast in linear networks.
216
Convergecast in Other Scenarios (cont.)
(a)
(b)
Multi-packet network.
Non-ideal radio propagation characteristics.
217
Conclusions
� A minimal time distributed convergecast scheduling algorithm
for sensor networks.
� The optimal convergecast schedule consists of
3N -3 timeslots, where N is the number of nodes in the network.
� More than 50% of the energy can be saved by using the
proposed sleep schedule.proposed sleep schedule.
218
Reference
� S. Gandham, Y. Zhang, and Q. Huang, “Distributed time-optimal scheduling
for convergecast in wireless sensor networks,” Computer Networks, vol. 52,
pp. 610-629, 2008.
219
Chapter 4.7
Data centric networkingData centric networking
220
Outline
� 4.7.1 Data centric routing
� 4.7.2 Data-centric storage
� 4.7.2.1 One-dimensional data storage
� 4.7.2.2 Multi-dimensional data storage
� 4.7.2.3 Hierarchical data storage
221
4.7.1 Data Centric Routing
� A central querier/data sink (or collection of queriers/sinks)
issues queries that sources in the network respond to. Due to
energy constraints it is desirable for much of the data
processing to be done in-network, and this has led to the
concept of data centric information routing, in which the
queries and responses are for named data.queries and responses are for named data.
� A sensor node is not an identity (address)
� Content based and data centric
� Where are nodes whose temperatures will exceed more than 10 degrees for
next 10 minutes?
� Tell me in what direction that vehicle in region Y is moving?
� Give me periodic reports about animal location in region A every 30 seconds.
222
Data Centric Routing (cont.)
� Depending on the applications, there are likely to be different
kinds of queries in these sensor networks.
� The types of queries can be categorized in many ways, for
example:
� Continuous queries, which result in extended data flows (e.g. “Report
the measured temperature for the next 7 days with a frequency of 1 the measured temperature for the next 7 days with a frequency of 1
measurement per hour”) versus One-shot queries, which have a simple
response (e.g. “Is the current temperature higher than 70 degrees?”)
� Aggregate queries, which require the aggregation of information from
several sources (e.g. “Report the calculated average temperature of all
nodes in region X”) versus Non-aggregate Queries which can be
responded to by a single node (e.g. “What is the temperature measured by
node x?”)
223
Data Centric Routing (cont.)
� Complex queries, which consist of several nested or batched sub-queries
(e.g. “What are the values of the following variables: X, Y, Z?”) versus
simple queries, which have no sub-queries (e.g. “What is the value of the
variable X?”)
� Queries for replicated data, In which the response to a given query can
be provided by many nodes (e.g. “Is there at least one target in the
area?”) and queries for unique data, in which the response to a given area?”) and queries for unique data, in which the response to a given
query can be provided only by one node.
224
Data Centric Routing (cont.)
� SPIN
� One-shot interactions
� 3-stage handshake protocol
� ADV – new data advertisement
� REQ – request for data
� DATA – data messageADV ADVREQDATA
225
ADVREQDATA
ADV
ADV
ADV
ADV
ADV
REQ
REQ
REQ
REQ
DATA
DATA
DATA
DATA
The SPIN protocol
Data Centric Routing (cont.)
� Directed Diffusion
� Repeated interactions
A simplified schematic for directed diffusion
226
4.7.2 Data-centric Storage
� Data centric storage
� Data is stored inside the network.
� All data with the same name (or data range) will be stored at the same
sensor network location
� E.g. an elephant sighting.
� Why data centric storage?� Why data centric storage?
� Energy efficiency
� Robustness against mobility and node failures
� Scalability
227
4.7.2
Data-centric storageData-centric storageOne-dimensional data storage
228
One-dimensional Data Storage
� Data-Centric Storage in Sensornets with GHT, a Geographic
Hash Table(GHT [Ratnasamy et al. 2003])
� Data Storage and Retrieval
� Perimeter Refresh Protocol
� Structured Replication
229
Data Storage and Retrieval
� GHT
� Put(k,v)-stores v (observed data)
according to the key k
� Get(k)-retrieve whatever value is
associated with key k
� Hash function
Hash the key in to the
(12,24)data
� Hash the key in to the
geographic coordinates
� Put() and Get() operations on the
same key “k” hash k to the same
location
user
queryresponse
Hash (‘elephant’)=(12,24)
Put (“elephant”, data) Get (“elephant”)
Hash (‘elephant’)=(12,24)
An example for GHT
230
Perimeter Refresh Protocol
� Assume key k hashes at
location L
� A is closest to L so it
becomes the home node
E
D
Replica
Replica
F
B
D
A
C
L
home
231
Structured Replication
� Augment event name with
hierarchy depth
� Given root r and given
hierarchy depth d
� Compute 4d – 1 mirror images
of r
(100, 100)(0, 100)
of r
(0, 0) (100, 0)
root point
level 1 mirror points
level 2 mirror points
Example of structured replication
with a 2-level decomposition
232
Conclusions
� Data centric storage entails naming of data and storing data at
nodes within the sensor network
� GHT uses Perimeter Refresh Protocol and structured
replication to enhance robustness and scalability
� DCS is useful in large sensor networks and there are many
detected events but not all event types are Queried detected events but not all event types are Queried
233
4.7.2
Data-centric storageData-centric storageMulti-dimensional data storage
234
Multi-dimensional Data Storage
� Multi-Dimensional Range Queries in Sensor Networks (DIM
[Li et al. 2003])
� Building Zones
� Data Insertion
� Query Propagation
235
Building Zones
� Divide network into zones.
� Each node mapped to one
zone.
� Encode zones based on
division.
� Each zone has a unique 6
5
1110
1111
43
21
1100111
0110
010
L∈∈∈∈[1/2, 1)L∈∈∈∈[0, 1/2)
T∈∈ ∈∈
[1/2
, 1
)
T∈∈ ∈∈
[3/4
, 1)
T∈∈ ∈∈
[1/2
, 3/4
)
code.
� Map m-d space to zones.
� Zones organized into a
virtual binary tree.
78
10
90001
0000 001 10
T∈∈ ∈∈
[0, 1
/2)
L∈∈∈∈[0, 1/4) L∈∈∈∈[1/4, 1/2) L∈∈∈∈[1/2, 3/4) L∈∈∈∈[3/4, 1)
T∈∈ ∈∈
[1/4
, 1/2
)T
∈∈ ∈∈[0
, 1/4
)
L: Light, T: Temperature
236
Data Insertion
� Encode events
� Compute geographic
destination
� Hand to GPSR
� Intermediate
nodes can refine
1110
1111
L∈∈∈∈[1/2, 1)
1
110
L∈∈∈∈[0, 1/2)
0111
0110
010
T∈∈ ∈∈
[1/2
, 1
)
T∈∈ ∈∈
[3/4
, 1)
T∈∈ ∈∈
[1/2
, 3/4
)
2
34
5
6
E1= <0.8, 0.7>
nodes can refine
the destination
estimation
L∈∈∈∈[0, 1/4)
0001
0000 001 10
T∈∈ ∈∈
[0, 1
/2)
L∈∈∈∈[1/4, 1/2) L∈∈∈∈[1/2, 3/4) L∈∈∈∈[3/4, 1)
T∈∈ ∈∈
[1/4
, 1/2
)T
∈∈ ∈∈[0
, 1/4
)
987
10
Store E1
L: Light, T: Temperature
237
Query Propagation
� Split a large query into smaller subqueries.
� Encode each subquery.
� Process subqueries separately, resolving locally or
1110
1111
L∈∈∈∈[1/2, 1)
1
110
L∈∈∈∈[0, 1/2)
0111
0110
010
T∈∈ ∈∈
[1/2
, 1
)
T∈∈ ∈∈
[3/4
, 1)
T∈∈ ∈∈
[1/2
, 3/4
)
2
34
5
6Q = <.75-1, .5-.75>
Q12= <.75-1, .75-1>Q11= <.5-.75, . 5-1>
locally or forwarding to other nodes based on their codes.
L∈∈∈∈[0, 1/4)
0001
0000 001 10
T∈∈ ∈∈
[0, 1
/2)
L∈∈∈∈[1/4, 1/2) L∈∈∈∈[1/2, 3/4) L∈∈∈∈[3/4, 1)
T∈∈ ∈∈
[1/4
, 1/2
)T
∈∈ ∈∈[0
, 1/4
)
987
10
Q10= <.75-1, .5-.75>
Q1= <0.5-1, 0.5-1>
L: Light, T: Temperature
238
Conclusions
� DIM resolves multi-dimensional range queries efficiently.
� Work that still needs to be done
� Skewed data distribution
� These can cause storage and transmission hotspots.
� Existential queries
� Whether there exists an event matching a multi-dimensional range.
� Node heterogeneity� Node heterogeneity
� Nodes with larger storage space assert larger-sized zones for themselves.
239
4.7.2
Data-centric storageData-centric storageHierarchical data storage
240
Hierarchical Data Storage
� Load balance and Efficient Hierarchical Data-Centric Storage
in Sensor Networks (HVGR [Zhao et al. 2008])
� Constructing the hierarchical architecture
� Storage load balancing
� Data Storage and retrieval
241
Constructing Hierarchical Architecture
� Assumptions
� Large
� Static
� density
•Each node knows the shortest root
•Each node knows the shortest
path to their first level landmark
242
Constructing Hierarchical Architecture (cont.)
� Assumptions
� Large
� Static
� density
•Each node knows the shortest
L2
L3
root
• Nodes know the shortest
path to their second level landmark
•Termination: the subregion only
contains the owner landmark and
the landmark’s one-hop neighbors.
•Each node knows the shortest
path to their first level landmark
L1
L4
L5
L4.1
L4.2
L4.3
243
Storage Load Balancing
N2N3
L2:[0.1,0.3)L3:[0.3,0.6)
N: the total number node in WSN
Ni: the number node in Ni’s subregion
N=50
0.1
0.2 0.3
N1
N4
N5
L1: [0,0.1)
L4:[0.6,0.8)L5:[0.8,1)
L4.1:[0.6,0.67)
L4.2:[0.67,0.73) L4.3:[0.73,0.8)
Node A table:
L1st:{L1:(0,0.1],L2 :(0.1,0.3],L3:(0.3,0.6],
L4:(0.6,0.8],L2 :(0.8,1]}
L2nd:{L4.1:(0.6,0.67], L4.2:(0.67,0.73],
L4.3:(0.73,0.8]}
L3rd:{L4.2.1: (0.67,0.73],
L4.2.2: (0.70,0.73]}
A
0.1
0.2
0.2
244
Data Storage and Retrieval
L1st:{L4 :(0.6,0.8],…}
Source(0.68)
Query (0.68)
L2 L3
Query (0.68)
L1st:{L4:(0.6,0.8],…}
L2nd:{L4.2:(0.67,0.73],…}
L1
L4L5
L4.1
L4.2L4.3
A
DL4.2.1
L4.2.2
L1st:{L4 :(0.6,0.8],…}
L2nd:{L4.2:(0.67,0.73],…}
L3rd:{L4.2.1: (0.67,0.73],L4.2.2: (0.70,0.73]}
245
Conclusions
� HVGR is very scalable, as the initialization overhead and
routing table size of each node is O(logN).
� HVGR design a simple hash mechanism so that HVGR can
provide a well load balanced data-centric storage system.
246
References� S. Ratnasamy, B. Karp, S. Shenker, D. Estrin, R. Govindan, L. Yin,
and F. Yu, “Data-Centric Storage in Sensornets with GHT, A Geographic Hash Table,” in Journal of Mobile Network Applications, vol. 8, no. 4, pp. 427-442, 2003.
� X. Li, Y. J. Kim, R. Govidan, and W. Hong, "Multi-Dimensional Range Queries in Sensor Networks," in Proceedings of the 1st International Conference on Embedded Networked Sensor Systems (SenSys'03), Los Angeles, CA, USA, pp.63-75, Nov. 2003.International Conference on Embedded Networked Sensor Systems (SenSys'03), Los Angeles, CA, USA, pp.63-75, Nov. 2003.
� Y. Zhao, Y. Chen, and S. Ratnasamy, "Load Balanced and Efficient Hierarchical Data-Centric Storage in Sensor Networks," in Proceedings of the 5th Annual IEEE Communications Society Conference on Sensor, Mesh and Ad Hoc Communications and Networks, San Francisco, California, USA, pp.560-568, June 2008.
247
Chapter 4.8
ZigBeeZigBee
248
Outline
� 4.8 The ZigBee Standard
� 4.8.1 Zigbee frame format
� 4.8.2 The Network Layer
� 4.8.3 The Application Layer
249
Outline
� 4.8 The ZigBee Standard
� 4.8.1 Zigbee frame format
� 4.8.2 The Network Layer
� 4.8.2.1 Network Formation and Address Assignment
� 4.8.2.2 ZigBee Routing protocol
� 4.8.2.3 Route Discovery
� 4.8.3 The Application Layer� 4.8.3 The Application Layer
250
The ZigBee Standard
� ZigBee is a low cost, low power, low complexity, and low data rate wireless
communication technology at short range. Based on IEEE 802.15.4, it is
mainly used as a low data rate monitoring and controlling sensor network
Applications
802.15.4
Zigbee
Specification
Application Framework
Network & Security
Medium Access Control (MAC) Layer
Physical (PHY) Layer
Application
Zigbee stack
Hardware
251
Zigbee Frame Format
� General frame format
Octets: 2 2 2 1 1 Variable
Frame Control Destination Source Radius Sequence Frame Frame Control Destination Source
Address
Radius Sequence
Number
Frame
Payload
Routing Fields
NWK Header NWK
Payload
252
Zigbee Frame Format (cont.)
� General frame format
Frame control field
Frame type setting
253
Zigbee Frame Format (cont.)
� Command frame format
Octets: 2 1 Variable
Frame control Routing fields NWK command
identifier
NWK command
payload
NWK header NWK payload
Command frame format
NWK header NWK payload
Frame type
254
Zigbee Frame Format (cont.)
� RREQ command
RREQ command
payload format
Command options field
255
Zigbee Frame Format (cont.)
� RREP command
RREP command
payload format
256
Outline
� 4.8 The ZigBee Standard
� 4.8.1 Zigbee frame format
� 4.8.2 The Network Layer
� 4.8.2.1 Network Formation and Address Assignment
� 4.8.2.2 ZigBee Routing protocol
� 4.8.2.3 Route Discovery
� 4.8.3 The Application Layer� 4.8.3 The Application Layer
257
The Network Layer
� ZigBee identifies three device types
� The ZigBee coordinator (one in the network) is an FFD managing the
whole network
� A ZigBee router is an FFD with routing capabilities
� A ZigBee end-device corresponds to a RFD or FFD acting as a simple
device
� The ZigBee network layer supports three types of network
configurations:
� Star topology
� Tree topology
� Mesh topology
258
The Network Layer (cont.)
ZigBee coordinator ZigBee router ZigBee end device
(a) Star network (b) Tree network (c) Mesh network
259
Outline
� 4.8 The ZigBee Standard
� 4.8.1 Zigbee frame format
� 4.8.2 The Network Layer
� 4.8.2.1 Network Formation and Address Assignment
� 4.8.2.2 ZigBee Routing protocol
� 4.8.2.3 Route Discovery
� 4.8.3 The Application Layer� 4.8.3 The Application Layer
260
Network Formation and Address Assignment
� Before forming a network, the coordinator determines
� Maximum number of children of a router (Cm)
� Maximum number of child routers of a router (Rm)
� Depth of the network (Lm)
� Note that a child of a router can be a router or an end device, so � Note that a child of a router can be a router or an end device, so
Cm ≥ Rm
� The coordinator and routers can each have at most Rm child
routers and at least Cm − Rm child end devices
261
Network Formation and Address Assignment
(cont.)
� For the coordinator, the whole address space is logically
partitioned into Rm + 1 blocks
� The first Rm blocks are to be assigned to the coordinator’s
child routers and the last block is reserved for the coordinator’s
own child end devices
� From Cm, Rm, and Lm, each router computes a parameter � From Cm, Rm, and Lm, each router computes a parameter
called Cskip to derive the starting addresses of its children’s
address pools
( )( )
1
1 1 ,if 1
1Otherwise,
1
Lm d
Cm Lm dRm
Cskip d Cm Rm Cm Rm
Rm
− −
+ × − −=
= + − − ×
−
262
Network Formation and Address Assignment
(cont.)
� The coordinator is said to be at depth d = 0, and d is increased
by one after each level
� Address assignment begins from the ZigBee coordinator by
assigning address 0 to itself
� If a parent node at depth d has an address Aparent , the n-th child
router is assigned to address router is assigned to address
� Aparent + (n − 1) × Cskip(d) + 1
� n-th child end device is assigned to address
� Aparent + Rm × Cskip(d) + n
263
Network Formation and Address Assignment
(cont.)
Addr = 12
Addr = 8
Addr = 9
Addr = 10
Cm = 5
Rm = 4
Lm = 2
A2
Addr = 7
Cskip = 1
ZigBee coordinator ZigBee router ZigBee end device
B1
Addr = 25
Addr = 0
Cskip = 6
A4
Addr = 19
Cskip = 1
A3
Addr = 13
Cskip = 1
Addr = 24
A1
Addr = 1
Cskip = 1
Addr = 6
Addr = 3
Addr = 2
264
Outline
� 4.8 The ZigBee Standard
� 4.8.1 Zigbee frame format
� 4.8.2 The Network Layer
� 4.8.2.1 Network Formation and Address Assignment
� 4.8.2.2 ZigBee Routing protocol
� 4.8.2.3 Route Discovery
� 4.8.3 The Application Layer� 4.8.3 The Application Layer
265
ZigBee Routing protocol
� In a ZigBee network, the coordinator and routers can directly
transmit packets along the tree
� When a device receives a packet, it first checks if it is the
destination or one of its child end devices is the destination
� If so, this device will accept the packet or forward this packet
to the designated child. Otherwise, it forwards the packet to its
parent
266
ZigBee Routing protocol (cont.)
� Assume that the depth of this device is d and its address is A.
This packet is for one of its descendants if the destination
address Adest satisfies A < Adest < A+ Cskip(d − 1), and this
packet will be relayed to the child router with address
( )( )
11
destA AA A Cskip d
− += + + ×
� If the destination is not a descendant of this device, this packet
will be forwarded to its parent
( )
( )( )
11
dest
r
A AA A Cskip d
Cskip d
− += + + ×
267
ZigBee Routing protocol (cont.)
Cm = 6
Rm = 4
Lm = 3
Addr = 125
Addr = 30
Addr = 126
Addr = 92
Addr = 63
Cskip = 7
Addr = 64
Cskip = 1
Addr = 1
( )
( )( )
11
dest
r
A AA A Cskip d
Cskip d
− += + + ×
ZigBee coordinator ZigBee router ZigBee end device
Addr = 31
Addr = 38
Addr = 1
Cskip = 7
Addr = 32
Cskip = 7Addr = 33
Cskip = 1
Addr = 40
Cskip = 1
Z
? ?
( )( )1A A Cskip d
Cskip d= + + ×
BC
A < Adest < A+ Cskip(d − 1)
A
268
Outline
� 4.8 The ZigBee Standard
� 4.8.1 Zigbee frame format
� 4.8.2 The Network Layer
� 4.8.2.1 Network Formation and Address Assignment
� 4.8.2.2 ZigBee Routing protocol
� 4.8.2.3 Route Discovery
� 4.8.3 The Application Layer� 4.8.3 The Application Layer
269
Route Discovery
Field Name Description
Destination Address 16-bit network address of the destination
Next-hop Address 16-bit network address of next hop towards destination
Entry Status One of Active, Discovery or InactiveEntry Status One of Active, Discovery or Inactive
Routing Table in ZigBee
270
Route Discovery (cont.)
Field Name Description
RREQ ID
(route request)
Unique ID (sequence number) given to every RREQ
message being broadcasted
Source Address Network address of the initiator of the route request
Sender Address Network address of the device that sent the most recent
lowest cost RREQ
Forward Cost The accumulated path cost from the RREQ originator to
the current device
Residual Cost The accumulated path cost from the current device to the
RREQ destination
Route Discovery Table
271
Route Discovery (cont.)
RREQ message
Create RDT entry and
record fwd path cost
RDT entry
exists for this
RREQ ?
NoYes
Is
RREQ for local
node or one of
end-device
children ?
Create RT entry
(Discovery_Underway)
And rebroadcast RREQ
Update RDT entry with
better fwd path cost
Does
RREQ report
A better fwd
path cost ?
Send RREPDrop RREQ
Yes
No
Yes
No
The RREQ processing
272
Route Discovery (cont.)
A
C
BDiscard route
request
route reply
S
D
T
Unicast
Broadcast
Without routing capacity
273
Outline
� 4.8 The ZigBee Standard
� 4.8.1 Zigbee frame format
� 4.8.2 The Network Layer
� 4.8.2.1 Network Formation and Address Assignment
� 4.8.2.2 ZigBee Routing protocol
� 4.8.2.3 Route Discovery
� 4.8.3 The Application Layer� 4.8.3 The Application Layer
274
The Application Layer
� A ZigBee application consists of a set of Application Objects (APOs) spread over
several nodes in the network
� The ZigBee Device Object (ZDO) is a special object which offers services to the
APOs
� The Application Sub layer (APS) provides data transfer services for the APOs and
the ZDO
275
References� P. Baronti, P. Pillai, V. Chook, S. Chessa, and F. Gotta, A. andFun Hu. Wireless sensor
networks: a survey on the state of the art and the 802.15.4 and zigbee standards.
Communication Research Centre, UK, May 2006.
� J. Bruck, J. Gao and A. A. Jiang, “MAP: Medial Axis Based Geometric Routing in Sensor
Network,” in Proceedings of ACM MobiCom, 2005.
� Q. Fang, J. Gao, L. Guibas, V. de Silva, and L. Zhang. GLIDER: Gradient landmark-based
distributed routing for sensor networks. In Proc. of the 24th Conference of the IEEE
Communication Society (INFOCOM’05), March 2005.
� B. Chen, K. Jamieson, H. Balakrishnan, and R. Morris. Span: An energy-efficient coordination
algorithm for topology maintenance in ad hoc wireless networks. In International Conference
on Mobile Computing and Networking (MobiCom 2001), pages 85–96, Rome, Italy, July 2001.
� Y. Xu, J. Heidemann, and D. Estrin. Geography-informed energy conservation for ad hoc
routing. In Proceedings of the ACM/IEEE International Conference on Mobile Computing and
Networking, pages 70–84, Rome, Italy, July 2001.
276
Conclusions
� Routing in sensor networks is a new area of research, with a
limited but rapidly growing set of research results
� We highlight the design trade-offs between energy and
communication overhead savings in some of the routing
paradigm, as well as the advantages and disadvantages of each
routing technique
277
routing technique
� Overall, the routing techniques are classified based on the
network structure into four categories: flat, hierarchical, and
location-based routing, and QoS based routing protocols.
� Although many of these routing techniques look promising,
there are still many challenges that need to be solved in sensor
networks