Data Communication
Localized Routing Metrics
Energy Efficiency
Energy is a very limited resource in WSNs
Energy efficiency is often a primary optimization goal
How to make data communication energy efficient?
• Apply data communication on an energy efficient topology;
example:
• Run MECN topology control first
• Apply greedy routing over MECN links only
• Incorporate energy efficiency in the protocol directly; example:
• Elaborate an energy aware greedy routing weight
• Apply greedy routing using this weight
Definition of energy minimizing localized routing metrics?
Using the Path Loss Formula
Remember: channel model for RF communication
Energy required to send a message from S to T amounts u(d) = d^a + c, while d = |ST|
Observation
• Assume an arbitrary number n of equidistant intermediate forwarding nodes can be placed between S and T
• Power required to send a message from S to T amounts: n * (d/n)^a + c) = n*(d/n)^a + n*c
• Define f(x)=x*(d/x)^a + x*c
• Is there an optimal x minimizing f(x)?
• If x0 or x ∞ then f(x)∞
• There exists one solution x0 which satisfies f’(x)=0
S T
d
Using the Path Loss Formula
n = floor(x
0) can be expressed in a closed form
• We can compute constant c1
• Which depends on a and c only and
• Which satisfies n = c1 * |ST|
Power consumption v(|ST|) in this case can be expressed in a closed form as well
• We can compute constant c2
• Which depends on a and c only and
• Which satisfies v(|ST|) = c2 * |ST|
How can we use this result to express a
localized routing metric?
Using the Path Loss Formula
Estimate on total power consumption when selecting next hop node A: u(|SA|) + v(|AD|)
Greedy routing: select the node in forward direction which minimizes the expression u(|SA|) + v(|AD|)
Result directly related to path loss formula d^a + c
What if other models are used?
• New theoretical analysis to compute u(.) and v(.) are necessary
• Problem if the model function can not be derived
• Problem if the power metric is given by empirical values
Other localized metric approaches A
S D
Assume minimal power consumption v(|AD|) on remaining path from A to D Assume power
consumption u(|SA|) to send a message
from S to A
The Cost over Progress Framework
Progress achieved by selecting node A: d-t
Assume each node provides same progress
Number of routing steps: d / (d-t)
Assume each routing step consumes energy u(s)
Approximation of total energy consumption: u(s) * (d / (d-t))
Greedy routing: select neighbor A which minimizes u(s) / (d-t)
Observe: any cost function can be plugged into this expression
S D
A t
d s
Increasing Network Lifetime
Define: network lifetime – time it takes until first node dies
Are energy optimized paths increasing network lifetime?
Observation
• Selecting minimum energy consuming path p(S,D) will only use nodes on this path
• If p(S,D) is used continuously, nodes along this path will die first
This motivates the use of cost metric c(v)
• The cost to use node amounts c(v) = 1/g(v)
• g(v) reflects v’s remaining power in [0,max_power]
Try to find a path p=v1… vn which reduces total cost
Local approximation: node s selects node v which minimizes c(v) / (|sd| -
|vd|)
) ( / 1 )
,
( v w g w
f ( p ) f ( v
i, v
i1)
c
Addressing Lifetime and Energy Consumption
Problem
• Addressing network lifetime only might produce very energy consuming paths
• Total energy consumption will affect network lifetime as well
Solution: combine cost c(w) metric and energy metric u(v,w) in one
Example: pc(v,w) = c(w) * u(v,w)
• If u(v,w) is high and another node x with about same or better c(w) exists it is unlikely that w is selected
• If c(w) is high and another node x with with about same or better u(v,w) exists it is unlikely that w is selected
V
W
X u(v,w)
u(v,x)
c(w)
c(x)
u(v,x) * c(x) u(v,w) * c(w)
Traffic Balance by Randomization
Requires criteria which selects some nodes in forward direction
• Example 1: All nodes closer to destination
• Example 2: All nodes lying a threshold closer to the destination
• Example 3: All nodes in forward direction but not exceeding a certain delay threshold
Randomly select a node out of this set
• Example 1: uniformly distributed
• Example 2: weighted distribution with highest weight on best node
Effect: traffic balance in a large relay area
• Reduced congestion
Data Communication
Beacon-less Routing
Beacon-less Routing (1)
Traditional greedy routing need information about all one- hop neighbors
• Periodic hello messages
• Transmitted with maximum signal strength
• Independently of current data traffic
• Problem of directional connections
Heissenbüttel, Brown: Beacon-less routing (BLR)
• Node is unaware of its neighbors
• Just broadcast a message to all unknown neighbors
• Receiving node introduces a small timeout before forwarding
• Node located at the “best” position introduces the smallest delay
• Nodes hearing of retransmission cancel the scheduled packet
Beacon-less Routing (2)
Problem: Message duplicates
• E and F are in backward direction
• E.g.: B introduces smallest delay
• A removes scheduled packet
• C does not hear transmission from B and forwards the packet too
Avoiding message duplicates
• Only nodes in a certain forwarding area allowed as candidate nodes
• Nodes in forwarding area are able to overhear retransmission of each other node in that area
Active selection method: control Message instead of full packet
• Forwarding node sends unicast to “winning”
node
• Large packet can be sent with reduced transmission power
S
D
A B
C
E F
Beacon-less Routing (3)
Possible delay functions (r=radius, p=progress, d=
distance)
• Basically MFR:
Max_delay(r-p)/r
• Slightly modified NFP:
Max_delay(p/r)
• An advanced delay function
Possible forwarding areas
• Circle: good forwarding area regarding progress and
successful hops
Data Communication
The Greedy Routing Failure
Greedy Routing Failure
Choosing node in backward direction may lead to packet loops
Nevertheless, there may exist a path from S to D (S may also be an intermediate node)
Loop-freedom and delivery rate are conflicting goals
Solutions?
Improved Single-Path Strategies
Improvements trying to reduce package drop probability
Example: GEDIR – Allow message to travel one hop in backward direction, i.e. packet dropped only if it would be sent back to the previous node
However, fact: greedy heuristic can not guarantee delivery
Additional requirements to provide delivery guarantees
• Network with specific properties supporting greedy routing
• Recovery strategies for greedy routing failures
Data Communication
Guaranteed Delivery Based on Memorization
Motivation
Many greedy routing schemes perform well in dense networks
Greedy routing has a small communication overhead
Desirable to run Greedy routing as long as possible
However, greedy routing might fail in sparse networks
Guaranteed delivery is desirable property as well
On the following slides: in case of failure run a recovery mechanism which requires memorizing past routing information
• In the message
• In the visited node
Recovery by Flooding
Stojmenovic, Lin: Partial flooding to guarantee delivery (f- GEDIR, f-MFR, f-DIR)
• Intermediate nodes handle packet according to GEDIR, MFR, …
• Concave node broadcasts packet to all neighbors
• To avoid message loops: concave node rejects further copies of the message, concave nodes are removed from the list of candidate nodes
Example: Message from S to D
S C
E
F
A
B G
H
D
unicast broadcast
Recovery by Flooding
When there is a path from source to destination then one of the neighbors lies on the path → guaranteed delivery
Observation: flooding produces many redundant message transmissions
Improvement: Component routing
• Connected components in node v – partitions in the one-hop neighborhood graph N(v) when removing v
• Algorithm
• Concave node determines connected components
• Forward message to only the best neighbor in each component
• Number of message transmissions reduced significantly
e.g. concave node has at most four connected components in the unit disk graph model
DFS-Based Routing
Jain et al.: Geographic Routing Algorithm (GRA)
• Intermediate node handles message greedily
• Concave node maintains route to destination node
• Start route discovery for outdated routing tables
• Stuck packet is routed to destination after successful route discovery
How to perform route discovery?
DFS-Based Routing
Depth first search from concave node S
• Yields an acyclic path from S to D
• Node X puts its address on route discovery packet p
• Forward to neighbor who has not seen p before
• Select neighbor Y which minimizes |XY|+|YD|
• If no possible neighbor exists, remove address from p and send it back to the node from which p was originally received
Alternative implementation: memorize DFS data in nodes
Other metrics may be applied on next neighbor selection
• Quality-of-service paths (delay and bandwidth criteria, connection time, …)
X W
1 2 3
Y1
Y2 root Y3
t1, t6: path = …W
t2: path = …W X
t3: path = …W X Y1
t4: path = …W X Y2 t5: path = …W X Y3
MEMORYLESS MESSAGE DELIVERY WITH
GUARANTEED DELIVERY?
67
68 Geographic Greedy Routing
T S
A
(a)
F B
D C
E
Strategy: select from nodes closer to the destination the one which minimizes a local cost metric
B
E T S
(b)
?
A
C
D F
G
Problem: greedy routing failure
69 Recovery based on Planar-Graph Routing
source node
destination node
70 Planar Graph Routing Example
T S
P
F
71 Planar Graph Routing Example
T S
Q
F
P
72 Planar Graph Routing Example
T S
F
P
73 Planar Graph Routing Example
T S
P
F
Übersicht
Untersuchte Netzstruktur und Problemstellungen
Topologie-basierte Routingprotokolle
Geographische Routingprotokolle
• Greedy-Routing und Planar-Graph-Routing
• Konstruktion von planaren Graphen
• Lokales Multicasting
• Beispiel MSTEAM
74
75 We need a Planar Graph
U V U V
Gabriel Graph (GG) Relative Neighborhood Graph (RNG)
W
U
V
Delaunay
Triangulation (DT)
76 We need a UDG or QUDG
UDG: Localized GG and RNG versions based on 1-hop neighbors
UDG: Localized DT version based on 2-hop neighbors (and less)
QUDG: Sort of localized GG (possibly the same for RNG)
Quasi unit disk graph (QUDG)
U V
Unit disk graph (UDG)
rmin rmax
U
77 Problems and Limitations
Locally constructing a planar graph in arbitrary networks is impossible
Even worse: localized unicast routing is impossible in arbitrary graphs
Localized single path algorithms deviation from shortest paths
• Let k be the hop/Euclidean length of the shortest path connecting s and t
• Localized single path algorithms may produce paths of length O(k2)
• Some even worse but some exist which are upper bounded by O(k2) u
x v
y
t s
78 Localized Unicast Routing in Practice!
Wireless network graph has structure!
Aim at localized unicast approaches with high delivery rate
Method of choice: sorts of clustering
arbitrary graph wireless network graph
79 Geographic Clustering
80 Geographic Clustering
81 Geographic Clustering
82 Geographic Clustering
83 K-Hop Clustering
84 K-Hop Clustering
85 K-Hop Clustering
86 K-Hop Clustering
Übersicht
Untersuchte Netzstruktur und Problemstellungen
Topologie-basierte Routingprotokolle
Geographische Routingprotokolle
• Greedy-Routing und Planar-Graph-Routing
• Konstruktion von planaren Graphen
• Lokales Multicasting
• Beispiel MSTEAM
87
88 Localized Multicast Forwarding Problem
Assumptions:
Location system
Nodes know position of
• Themselves
• Their neighbors
• The destinations source node
destination node destination node
destination node
destination node
89
Building Blocks
T1
T3
T2 S
B A
D C
90
Building Blocks – Message Split
T1
T3
T2 S
B A
D C
91
Building Blocks – Next Hop Selection
T1
T3
T2 S
B A
D C
?
92
Building Blocks – Recovery
T1
T3
T2 S
B A
D C
?
?
Übersicht
Untersuchte Netzstruktur und Problemstellungen
Topologie-basierte Routingprotokolle
Geographische Routingprotokolle
• Greedy-Routing und Planar-Graph-Routing
• Konstruktion von planaren Graphen
• Lokales Multicasting
• Beispiel MSTEAM
93
94 EMST Backbone Assisted Localized Routing
T9
T7
T6
T4
T1
T3 S
T8
T5
T2 T1,…,T9
95 EMST Backbone Assisted Localized Routing
T8 T9
T7
T6
T5
T4
T1
T2
T3 S
EMST(S,T1,…,T9)
96 EMST Backbone Assisted Localized Routing
T8 T9
T7
T6
T5
T4
T1
T2
T3
D1
D2 D3
S
97 EMST Backbone Assisted Localized Routing
T8 T9
T7
T6
T5
T4
T1
T2
T3 A
B
C
S
T7,T8,T9
T1,T2,T3 T4,T5,T6
98 The Cost over Progress Framework
T3
T1
W V
T2 S
Which one is the better next hop node?
T1,T2,T3
99 The Cost over Progress Framework
Approximate expected number of hops H(S,V)
H(S,V) |EMST(S,T1,T2,T3)| / (|EMST(S,T1,T2,T3)| - |EMST(V,T1,T2,T3)|)
Approximate expected cost C(S,V) = cost(S,V) * H(S,V)
Select node X which provides progress and minimizes C(S,X) T3
T1
W V
T2 S
100
MSTEAM & MFACE
S T6
T5
T4
T3
T1
T2
F1
F2
101
MSTEAM & MFACE
S T6
T5
T4
T3
T1
T2
F1
F2
102
MSTEAM & MFACE
S
U
V W
F1 F2
F3
T1
T2 T3
103
MSTEAM & MFACE
S
U
V W
F1 F2
F3
T1
T2 T3
104
MSTEAM & MFACE
S
U
V W
F1 F2
F3
T1
T2 T3
105
MSTEAM & MFACE
S
U
V W
F1 F2
F3
T1
T2 T3 p
106
MSTEAM & MFACE
S
U
V W
F1 F2
F3
T1
T2
107
MSTEAM & MFACE
S
U
V W
F1 F2
F3
T2
Zusammenfassung
Untersuchte Netzstruktur und Problemstellungen
Topologie-basierte Routingprotokolle
Geographische Routingprotokolle
108
Zusammenfassung
Wir betrachteten hier: drahtlose Vernetzung ohne aufwendige (und ggf.
kostenpflichtige) Infrastruktur
Abdeckung größerer Gebiete trotz limitierter Kommunikationsreichweite Multihop-Kommunikation
Wir haben es hier somit hauptsächlich mit einem Netzwerkproblem zu tun
Wesentliche Probleme: Routing und Topologiekontrolle
Anpassung traditioneller Routing-Verfahren: Topologie-basiertes Routing
Neuer Routing-Ansatz auf der Basis von Knotenkoordinaten
Dieser Ansatz erlaubt ganz neue Formen der Datenkommunikation und generell ganz neue Formen von Netzorganisation
Generelles Paradigma, um mit der Dynamik solcher infrastrukturlosen Multihop- Netze umzugehen: lokale Algorithmen/Verfahren
Dieses Paradigma ist auch zur Beherrschung von komplexen und dynamischen Internet-Overlay-Topologien anwendbar
Mit den hier behandelten lokalen Verfahren wurde nur ein kleiner Ausschnitt eines interessanten Forschungsfeldes betrachtet
Mehr in der Vorlesung „Lokale Netzstrukturen“
109