WIRELESS MESH NETWORKS

Ian F. AKYILDIZ* and Xudong WANG**

* Georgia Institute of Technology

BWN (Broadband Wireless Networking) Lab &

** TeraNovi Technologies
4. ROUTING PROTOCOLS

2
 Features for Optimal Routing in WMNs

◼ Scalability

◼ Fast route discovery and rediscovery (for reliability)

◼ Mobile user support (for seamless and efficient handover)

3
 Features for Optimal Routing in WMNs

▪ Multiple Performance Metrics

minimum hop-count → ineffective!!
e.g., link quality and round trip time (RTT)

▪ Robustness
(link failures or congestions, fault-tolerant, load balancing)

▪ Adaptive Support of Both Mesh Routers and Mesh Clients

(support mobility and power efficiency)

4
 Features for Optimal Routing in WMNs

◼ Flexibility
– Work with/without gateways, different topologies

◼ QoS Support
– Consider routes satisfying specified criteria

◼ Multicast
– Important for some applications (e.g., emergency response)

5
 Prelim Routing Protocols for WMNs

* Apply routing algorithms derived for Ad Hoc Networks

6
CLASSIFICATION OF ROUTING PROTOCOLS
FOR AD HOC NETWORKS

7
 OVERVIEW
◼ Flat
– Reactive
⚫ DSR – Dynamic Source Routing
⚫ AODV – Ad hoc On demand Distance Vector
– Proactive
⚫ FSR – Fisheye State Routing
⚫ FSLS – Fuzzy Sighted Link State
⚫ OLSR – Optimized Link State Routing Protocol
⚫ TBRPF – Topology Broadcast Based on Reverse Path Forwarding

8
 OVERVIEW

◼ Hierarchical
– CGSR – Clusterhead-Gateway Switch Routing
– HSR – Hierarchical State Routing
– LANMAR – Landmark Ad Hoc Routing
– ZRP – Zone Routing Protocol

9
 OVERVIEW

◼ Geographical Routing
– DREAM – Distance Routing Effect Algorithm for Mobility
– GeoCast – Geographic Addressing and Routing
– GPSR – Greedy Perimeter Stateless Routing
– LAR – Location-Aided Routing

10
 Reminder: Ad Hoc Routing Protocols

◼ Proactive Protocols (Actively seeks for routes/paths)

– Determine routes independent of traffic pattern
– Traditional link-state and distance-vector routing protocols are
proactive

◼ Reactive Protocols (Seek for routes/paths only when required)

– Maintain routes only if needed

◼ Hierarchical Routing
– Introduces hierarchy to the flat network

◼ Geographic Position Assisted Routing

– Why send packets North when the destination is South?
11
 Routing Protocols from Ad Hoc Networks Used for WMNs

– Dynamic Source Routing (DSR)

in Microsoft mesh networks

D.B. Johnson, D.A. Maltz, and Y.-C. Hu,

“The dynamic source routing protocol for mobile ad hoc networks (DSR),”
IETF Internet-Draft, 2004.

12
 Routing Protocols from Ad Hoc Networks Used for WMNs

– AODV (ad-hoc on-demand distance vector) routing

⚫ Used by many other companies
⚫ Major building block for the routing framework of IEEE
802.11s

C. E. Perkins, E. M. Belding-Royer, I. D. Chakeres, “Ad hoc On-Demand Distance

Vector (AODV) Routing”, IETF Draft, Jan. 2004.

IEEE 802.11s Task Group, “Joint SEE-Mesh/Wi-Mesh proposal to 802.11

TGs overview,” IEEE Doc: 802.11-05/0567r6, 2006.

13
 Routing Protocols from Ad Hoc Networks Used for WMNs

– Topology broadcast based on reverse path forwarding

(TBRPF) protocol in Firetide Networks

R. Ogier, F. Templin, and M. Lewis, “Topology dissemination based on

reverse-path forwarding (TBRPF),” IETF RFC 3684, 2004.

14
 Dynamic Source Routing (DSR)
D. B. Johnson, D. A. Maltz, Y.-C. Hu,
“Dynamic Source Routing Protocol for Mobile Ad Hoc Networks”,
IETF Draft, April 2004.

◼Based on Source Routing principle!

◼On-demand

◼ Route computation performed on a per-connection basis

15
 Dynamic Source Routing (DSR)

◼ Source, after route computation, appends each packet with a

source-route information

◼ Intermediate hosts forward packet based on source route

◼ TWO PHASES: ROUTE DISCOVERY &

ROUTE MAINTENANCE

16
 Dynamic Source Routing (DSR):
ROUTE DISCOVERY

◼ When node S wants to send a packet to node D, but does

not know a route to D, node S initiates a Route Discovery

◼ Source node S floods (broadcasts) Route Request (RREQ) packet.

◼ RREQ packet contains

* DESTINATION ADDRESS
* SOURCE NODE ADDRESS and
* A UNIQUE IDENTIFICATION NUMBER.

17
 Dynamic Source Routing (DSR):
ROUTE DISCOVERY

◼If the node is the receiver (i.e., has the correct

destination address) then returns the packet to the
sender

◼If the packet has already been received earlier

(identified via ID) then discard the packet

18
 Dynamic Source Routing (DSR):
ROUTE DISCOVERY

◼ Each node receiving the packet checks whether

it knows of a route to that destination.

◼ If it does not, it appends/adds its own

identifier (address) to the route record and
forwards the RREQ packet.
19
Dynamic Source Routing (DSR):
ROUTE DISCOVERY
Y

Z
S E
F
B
C M L
J
A G
H
K
D
I N

20
 Dynamic Source Routing (DSR):
ROUTE DISCOVERY
Y
Broadcast transmission
[S] Z
S E
F
B
C M L
J
A G
H D
K
I N

Represents transmission of RREQ

[X,Y] Represents list of identifiers appended to RREQ
21
 Dynamic Source Routing (DSR):
ROUTE DISCOVERY
Y

Z
S E
[S,E]

F
B
C M L
J
A [S,C] G
H D
K
I N

Node H receives packet RREQ from two neighbors:

Potential for collision
22
 Dynamic Source Routing (DSR):
ROUTE DISCOVERY
Y

Z
S E
F [S,E,F]
B
C M L
J
A G
H D
[S,C,G] K
I N

Node C receives RREQ from G and H, but does not forward

it again, because node C has already forwarded RREQ once
23
 Dynamic Source Routing (DSR):
ROUTE DISCOVERY
Y

Z
S E
F [S,E,F,J]
B
C M L
J
A G
H D
K
I [S,C,G,K] N

• Nodes J and K both broadcast RREQ to node D

• Since nodes J and K are hidden from each other, their transmissions
may collide
24
 Dynamic Source Routing (DSR):
ROUTE DISCOVERY
Y

Z
S E
[S,E,F,J,M]
F
B
C M L
J
A G
H
K
D
I N

• Node D does not forward RREQ, because node D is the intended

target of the route discovery

25
 Route/Path Discovery in DSR

◼Destination D on receiving the first RREQ, sends a

Route Reply (RREP)

◼RREP is sent on a route obtained by reversing the

route appended to received RREQ

◼RREP includes the route from S to D on which RREQ

was received by node D
26
Route/Path Discovery in DSR
Y

Z
S RREP [S,E,F,J,D]
E
F
B
C M L
J
A G
H
K
D
I N

Represents RREP control message

27
 DATA DELIVERY IN DSR

◼ Node S on receiving RREP, caches the route included in the RREP

◼ When node S sends a data packet to D, the entire route is

included in the packet header
– hence the name Source Routing

◼ Intermediate nodes use the Source Route included in a

packet to determine to whom a packet should be forwarded

28
 Data Delivery in DSR
Y

DATA [S,E,F,J,D] Z
S E
F
B
C M L
J
A G
H D
K
I N

Packet header size grows with route length

29
 DSR Optimization: Route Caching

◼Each node caches a new route it learns by any means

◼When node S finds route [S,E,F,J,D] to node D, node

S also learns route [S,E,F] to node F

◼When node K receives Route Request [S,C,G] destined

for K, it learns route [K,G,C,S] to node S

30
 DSR Optimization: Route Caching

◼When node F forwards RREP [S,E,F,J,D],

node F learns route [F,J,D] to node D

◼When node E forwards Data [S,E,F,J,D] it learns

route [E,F,J,D] to node D

◼A node may also learn a route when it overhears Data

packets
31
 Advantages of Use of Route Caching

– can speed up route discovery

– can reduce propagation of route requests

32
 Use of Route Caching
[S,E,F,J,D]
[E,F,J,D]
S E [F,J,D],[F,E,S]

B F [J,F,E,S]
C M L
J
A [C,S] G
H [G,C,S] D
K
I N

[x,y,w]Represents cached route at a node

(DSR maintains the cached routes in a tree format)
33
 Use of Route Caching:
Can Speed up Route Discovery
[S,E,F,J,D]
[E,F,J,D]
S E [F,J,D],[F,E,S]
F [J,F,E,S]
B
C [G,C,S] M L
J
A [C,S] G
H D
[K,G,C,S] K
I RREP N
When node Z sends a route request RREQ RREQ
for node C, node K sends back a route Z
reply [Z,K,G,C] to node Z using a locally
cached route
34
 Use of Route Caching:
Can Reduce Propagation of Route Requests

[S,E,F,J,D] Y
[E,F,J,D]
S E [F,J,D],[F,E,S]
F [J,F,E,S]
B
C [G,C,S] M L
J
A [C,S] G
H D
[K,G,C,S] K
I RREP N
RREQ
Z

Route Reply (RREP) from node K limits flooding of RREQ.

In general, the reduction may be less dramatic.
35
 Dynamic Source Routing:
Advantages
◼ Routes maintained only between nodes who need to
communicate
– reduces overhead of route maintenance

◼ Route caching can further reduce route discovery

overhead

◼ A single route discovery may yield many routes to the

destination, due to intermediate nodes replying from local
caches
36
 Dynamic Source Routing:
Disadvantages

◼Packet header size grows with route length

◼Flood of route requests may potentially reach all
nodes
◼Care must be taken to avoid collisions between
route requests propagated by neighboring nodes
– insertion of random delays before forwarding
RREQ

37
 Ad Hoc On-Demand Distance Vector Routing (AODV)
C. E. Perkins, E. M. Belding-Royer, I. D. Chakeres, “Ad hoc On-
Demand Distance Vector (AODV) Routing”, IETF Draft, Jan. 2004.

◼ AODV attempts to improve on DSR by maintaining routing

tables at the nodes, so that data packets do not have to
contain paths

◼ AODV retains the desirable feature of DSR that routes

are maintained only between nodes which need to
communicate

38
 AODV

– Hop-by-hop routing as opposed to source routing

– On-demand

– When a source node wants to send a message to some destination

node and does not already have a valid route to that destination,
it initiates a Path Discovery Process to locate the destination
(as in DSR case)

– It broadcasts the RREQ packet to its neighbors

39
 AODV
◼ RREQs are forwarded in a manner similar to DSR

◼ When a node re-broadcasts a RREQ, it sets up a reverse

path pointing towards the source
– AODV assumes symmetric (bi-directional) links

◼ When the intended destination receives a RREQ, it replies

by sending a Route Reply (RR)

◼ RR travels along the reverse path set-up when RREQ was

forwarded
40
 AODV

– When RREQ propagates, routing tables are updated at

intermediate nodes (for route to source of RREQ)

– When RREP is sent by destination, routing tables

updated at intermediate nodes (for route to
destination), and propagated back to source

41
 AODV

– Each node maintains its own sequence number

and a broadcast ID.

– The broadcast ID is incremented for every

RREQ the node initiates

42
 AODV

– The node’s IP address and the broadcast ID

uniquely identify a RREQ.

– Along with its own sequence number and broadcast

ID, the source node includes in the RREQ the most
recent sequence number it has for the destination.

43
 AODV

– Intermediate nodes can reply to the RREQ only if

they have a route to the destination whose
corresponding Destination Sequence Number is
greater than or equal to that contained in the RREQ.

– During the process of forwarding the RREQ,

intermediate nodes recording their route tables with the
address of the neighbor from which the first copy
of the broadcast packet is received→ establishing a
reverse path.
44
 AODV

– If more same RREQs are received later, they are

discarded.

– RREP packet is sent back to the neighbors and the routing

tables are accordingly updated.

45
 Route Requests in AODV
Y

Z
S E
B F
C M L
J
A G
H D
K
I N

46
 Route Requests in AODV
Y
Broadcast transmission
Z
S E
F
B
C M L
J
A G
H D
K
I N

Represents transmission of RREQ

47
 Route Requests in AODV
Y

Z
S E
F
B
C M L
J
A G
H D
K
I N

Represents links on Reverse Path

48
 Reverse Path Setup in AODV
Y

Z
S E
F
B
C M L
J
A G
H D
K
I N

• Node C receives RREQ from G and H, but does not forward

it again, because node C has already forwarded RREQ once
49
Reverse Path Setup in AODV
Y

Z
S E
F
B
C M L
J
A G
H D
K
I N

50
 Reverse Path Setup in AODV
Y

Z
S E
F
B
C M L
J
A G
H D
K
I N

• Node D does not forward RREQ, because node D

is the intended target of the RREQ

51
 Route Reply in AODV
Y

Z
S E
F
B
C M L
J
A G
H D
K
I N

Represents links on path taken by RREP

52
Forward Path Setup in AODV
Y

Z
S E
F
B
C M L
J
A G
H D
K
I N
Forward links are setup when RREP travels along
the reverse path

Represents a link on the forward path

53
 Data Delivery in AODV
Y
DATA
Z
S E
F
B
C M L
J
A G
H D
K
I N

Routing table entries used to forward data packet.

NOTE: Route is not included in packet header as in DSR.

54
 Routing Protocols for WMNs

* Special considerations

- WMN routers differ from MANET routers

* Power supply
* Mobility
* Separation of WMN routers and clients

55
 Routing Challenges in WMNs

◼ Routing in WMNs is much more complicated than in Ad Hoc Networks

REASONS:
1) Network topology is variable and inconsistent (same as ad hoc
networks)

2) Depending on the performance goal in routing, it may not be

possible to determine a routing path solely based on network topology

3) There may not be an optimal solution for a given routing problem

4) Routing traverses both mesh routers and mesh clients that have
different networking capabilities
56
 Design Principles

1) Maintaining a consistent and stable network topology

2) Performing dynamic and adaptive routing

3) Developing new routing metrics

4) Considering tradeoff between cross-layer design and

single-layer solution

57
 Design Principles

5) Deriving distributed algorithms for routing

6) Ensuring scalability in routing

7) Adaptively supporting both mesh routers and mesh clients

58
 PERFORMANCE METRICS

Per-Flow Parameters:
(e.g., delay, packet loss ratio, and delay jitter and other parameters
such as hop-count, per-flow throughput, and intra-flow interference)

Per-Node Parameters:
(computational complexity and power efficiency)

Per-Link Parameters:
(e.g., link quality, channel utilization, transmission rate, and congestion)

Inter-Flow Parameters:
(e.g., inter-flow interference and fairness)

Network-Wide Parameters:
(e..g., total throughput or total delay)

59
 OVERVIEW OF ROUTING METRICS

* Hop-Count

* Per-Hop RTT

* Per-Hop Packet Pair Delay

* Expected Transmission Count (ETX)

60
 OVERVIEW OF ROUTING METRICS

* Expected Transmission on a Path (ETOP)

* Expected Transmission Time (ETT)

* Weighted Cumulative ETT (WCETT)

* Bottleneck Link Capacity (BLC)

* Expected Data Rate (EDR)

* Airtime Cost Routing Metric

61
 Hop Count

* Minimum hop counting (the link quality is binary)

* Simple and requires no measurements

Disadvantages:
* Can lead to poor throughput
* Link quality → all links do not have the same quality
* Does not take packet loss or bandwidth into account
* Route that minimizes hop count does not
necessarily maximize the throughput

62
 Per-Hop RTT
Adya A, Bahl P, Padhye J, Wolman A and Zhou L,
“A multi-radio unification protocol for IEEE 802.11 wireless networks”,
Proc. IEEE BroadNets 2004

◼ Measured by sending unicast probe packets between

neighboring nodes

◼ Then calculating the time spent on the probe-ack procedure

◼ A weighted moving average is needed to get a smoother

measurement, because one sample cannot really reflect the
actual link status.

63
 Per-Hop RTT

◼ Based on per-hop RTT, a routing protocol selects a routing

path with the least sum of RTTs of all links on the path.

◼ Per-hop RTT is able to capture

* the packet loss ratio in a link
* the traffic load
* queuing delay in two nodes on the link, and
* contention status in all neighboring nodes.

64
 Per-Hop RTT

* Loss will cause RTT to increase due to ARQ

* If ARQ fails, RTT is increased by some percentage

* This metric is load dependent

* Channel contention increases RTT

65
 Per-Hop RTT
Kim K-H and Shin KG, “On accurate measurement of link quality in
multi-hop wireless mesh networks,” Proc. ACM MOBICOM 2006

Its effectiveness is constrained by two problems:

PROBLEM 1:
Per-hop RTT is too much dependent on the traffic load/queuing delay, which interferes
with the accuracy of per-hop RTT and thus, can easily lead to route instability.

If a separate queue is assigned to probe packets, then it can accurately measure the
link quality but cannot reflect the traffic load.

A solution to this problem is to adopt a link measurement

scheme !
66
 Per-Hop RTT
PROBLEM 2:
◼ Accuracy of per-hop RTT measurement totally relies on the weighted
moving average scheme.

◼ For large variations in measurements cause unreliable values for per-

hop RTT
(no matter what weight is applied in the weighted moving average scheme.)

◼ Overhead of the probe-ack procedure

REMARK:
Per-hop RTT captures per-link performance parameters, although
measurement is actually carried out at the network layer.

67
 Disadvantages of Per-Hop RTT

Disadvantages:
* Does not take link data rate into account.
* High overhead.
* Load dependent metric may cause route flaps
* Need to insert probe at head of interface queue to
avoid queuing delay
* Not scalable - every pair needs to probe each other

68
 Per-Hop Packet Pair Delay
Draves R, Padhye J and Zill B, “Routing in multi-radio, multi-hop
wireless mesh networks,” Proc. ACM MOBICOM 2004.

◼ Measured by sending two back-to-back probe packets from a node to its neighbor

◼ First → a small probe packet; Second → large

◼ When the neighbor receives these two packets, it finds the delay in-between
them and then sends such information back to the probing node

◼ Since relative delay is used to measure the per-hop delay, per-hop PPD
measurement is less impacted by queueing delays or traffic load in a node

69
 Per-Hop Packet Pair Delay

* However, impact by traffic load still exists, because whether or

not being able to send probe packets in a link of two nodes also
depends on the queuing delays of other neighboring nodes.

EXAMPLE:
when Node A sends a probe packet to B, if A’s neighbor C is also
sending a very high traffic load to A, then A has to delay its probe to B.

Therefore, per-hop packet pair delay still has to capture

the route instability issue.

70
 Per-Hop Packet Pair Delay

* Large overhead than per-hop RTT, due to the need of more

probe packets

◼ Its performance is also dependent on the weighted moving

average scheme, and assumes the variation of measurements
are small.

◼ Similar to per-hop RTT, per-hop PPD only captures per-link

performance parameters.

71
 Expected Transmission Count (ETX)
De Couto DSJ, Aguayo D, Bicket J, and Morris R,
“A high-throughput path metric for multihop wireless routing,”
in Proc. ACM MOBICOM 2003

◼ ETX of a link is the expected number of transmissions before a packet

is successfully delivered on a link
◼ For a route, the ETX is the sum of the ETXs on all links
◼ Captures the link quality and packet loss on both directions of a link
◼ The route ETX can detect interference among links of the same route
→ the larger the route ETX, the more self-interference on the route.

72
 Expected Transmission Count (ETX)

◼ Every period of  seconds a node sends a broadcast probe message

to all its neighbors

◼ Each neighbor records the number of received probe messages

(denoted by nw) during a period of w seconds, where w > 

◼ Thus, the delivery ratio of sending a packet from the probing node
to its neighbor is:

nw
w /
73
 Expected Transmission Count (ETX)

◼ If a probing node embeds the information of nw from all its neighbors to

the probe packet

◼ Then each of its neighbors can derive the packet delivery ratio from the
neighbor to the probing node

◼ With the delivery ratio at both forward and reverse directions, denoted
by df and dr, respectively, ETX is calculated as:
1
ETX =
d f  dr

74
 Advantages of
Expected Transmission Count (ETX)

◼Lower overhead because broadcast rather than

unicast is applied to probe messages.

◼Does not measure delays, so the measurement based

on probe messages are not impacted by queueing
delays in a node.

75
 Disadvantages of
Expected Transmission Count (ETX)

– Probe messages experience different packet loss ratios than

unicast messages

because broadcast messages use more robust modulation and

coding schemes, and thus have low transmission rates

– ETX does not take into account the differences in packet size
for different traffic flows and the different capacities for
different links

76
 Disadvantages of
Expected Transmission Count (ETX)

– Estimation method in ETX may not be accurate

* It relies on the mean loss ratio;
* Wireless links usually experience bursty losses.

77
 SO FAR
* ETX metric performs best in static scenarios

* It is insensitive to load
* RTT is most sensitive to load
* Packet-Pair suffers from self-interference on multi-hop paths

* Minimum hop count based routing seems to perform best in mobile

scenarios

* Schemes based on measurements of link quality does not converge

quickly

78
 Expected Transmission on a Path (ETOP)

* When a routing path is selected in many routing protocols, the position

of a link is not considered in the routing metric

◼ This is true if the link layer allows an infinite number of

retransmissions, because a retransmitted packet has the same impact
on upper layer no matter at which link retransmission happens

◼ However, if the link layer allows only a limited number of

retransmissions, end-to-end retransmission has to be carried out.

79
 Expected Transmission on a Path (ETOP)
Jakllari G, Eidenbenz S, Hengartner N, Krishnamurthy S and Faloutsos M,
“Link positions matter: a non-commutative routing metric for wireless mesh
networks,” Proc. IEEE INFOCOM 2008

◼ Comparing two links, even if their ETX is the same, the one closer to
the destination can result in higher transport layer retransmissions,
i.e., this link can lead to worse performance if it would be selected.

◼ ETOP solves the above problem by taking into account the relative
position of a link on a routing path when the routing cost of the path is
calculated.

80
 Expected Transmission on a Path (ETOP)

◼ Consider a routing path with n links from node v 0 to node vn

→ its cost is denoted by Tn

◼ For a packet to be delivered end-to-end through this routing path,

the needed number of end-to-end attempts is assumed to be Yn.

◼ In an end-to-end attempt j, the number of links that a packet has

been traversed before it is dropped by the link layer is denoted as M

◼ The number of link layer transmissions at node j is assumed to be H j

.
81
 Expected Transmission on a Path (ETOP)

◼ ETOP of a routing path is the expectation of Tn

 n−2  n −1
E[Tn ] =  K +  (E[H j | H j  K ] P[M  j | M  n])  E[Yn − 1] +  E[H j | H j  K ]
 j=0  j=0

◼ Captures the total number of link layer transmissions of a given routing path under
all possible end-to-end attempts

◼ Compared to ETX, ETOP can improve transport layer throughput, because a routing
path is selected with a least number of overall link layer retransmissions.

82
 Expected Transmission Time (ETT) and
Weighted Cumulative ETT (WCETT)
Draves R, Padhye J and Zill B, “Comparisons of routing metrics for static
multi-hop wireless Networks,” in Proc. ACM SigComm, 2004

◼ Expected Transmission Time (ETT)

– An extended version of ETX.
– Based on ETX, ETT considers the impact of both packet size and
link quality
S
ETT = ETX S: packet size, B: link bandwidth.
B

– ETT reflects the expected packet transmission time on a link.

83
 Expected Transmission Time (ETT) and
Weighted Cumulative ETT (WCETT)

– For a routing path, the expected transmission time can be the sum
of ETTs of all links on the path.

– However, ETT does not take into account channel diversity in

WMNs using multiple radios at some nodes

– → To resolve this issue, a routing metric called WCETT is

proposed

84
 Expected Transmission Time (ETT) and
Weighted Cumulative ETT (WCETT)

◼ Weighted Cumulative ETT (WCETT)

n: number of hops on a routing path,

n
WCETT = (1 −  ) ETTi +  max X j
i =1 1 j  k k: # of available channels for multi-radio
operation.

so  max X j
n
X j =  ETT
HOP i on channel i
i 1 j  k
finds the bottleneck channel of a

given routing path.

– First term considers the overall expected transmission time of the routing path.
– Second term captures the transmission time on the bottleneck channels.
– In this way, WCETT takes into account the tradeoff between overall routing delay and
channel diversity utilization.
85
 Expected Transmission Time (ETT) and
Weighted Cumulative ETT (WCETT)

◼ ETT enhances the performance of ETX by mapping

packet size and link BW into the transmission time.

◼ However, it uses a similar estimation scheme as that of

ETX, so it has similar problems of ETX, i.e., inaccurate
estimation, bottleneck routes, etc.

86
 Expected Transmission Time (ETT) and
Weighted Cumulative ETT (WCETT)

WCETT is not applicable to WMNs based on single-radio

multi-channel operation for two reasons:

1) broadcast probe messages cannot be sent on different

channels of the same radio simultaneously;

2) the channel switching time can be comparable to ETT of a link.

87
 Bottleneck Link Capacity (BLC)
Liu T and Liao W, “Capacity-aware routing in multi-channel multi-rate
wireless mesh networks,” in Proc. IEEE ICC, 2006

◼ BLC is derived based on the expected busy time (EBT) of

transmitting a packet on a link.

◼ EBT can be estimated by considering the packet loss rate (PLR) and
transmission mechanism in the MAC layer.
– If RTS-CTS-Data-Ack handshake is used for packet transmission as in an
IEEE 802.11 MAC,

Thandshake Thandshake: Total transmission time of one RTS-CTS-Data-Ack

EBT = Ep: PLR.
1− ep

88
 Bottleneck Link Capacity (BLC)

◼ Based on EBT, a residual capacity of a link considered as

defined as the ratio between the idle time and EBT.
– Considering a path P, if the residual capacity of link i is LCi,
then BLC is given by

min LCi K: Length of the routing path P

BLC = iP
μ: fine-tuning parameter.
K

89
 Bottleneck Link Capacity (BLC)

◼ Residual capacity of the bottleneck link of a routing path

◼ Dividing the minimum residual capacity by a certain number

is for penalizing a long routing path

◼ Because busy time is considered in BLC, load-balancing in

links has been taken into account

90
 Bottleneck Link Capacity (BLC)

◼ However, the self-interference of a routing path is not considered,

as the minimum residual capacity is considered in BLC

– If two routing paths have different self-interferences, then the

bottleneck link can have the same residual capacity.

– The same problem applies to interference from other routing

paths.

91
 Expected Data Rate (EDR)

◼ EDR integrates the expected transmission count and expected transmission

contention degree (TCD) into the same routing metric

– TCD of a link is the time that is spent on retransmitting non-acknowledged packets

over a given period.

– Considering link k on a routing path, if the sum of TCDs of links that interfere link
k is Ik, then the the EDR of link k is

 Γ: maximum transmission rate of link k.

EDRk =
I k ETX k

◼ For EDR of a routing path, it is defined as the EDR of the bottleneck link.

92
 Problems of Expected Data Rate (EDR)

– EDR integrates two closely related parameters: ETX ad TCD.

– In fact, given the same packet length, if ETX is large, TCD

is large too.

→ why ETX and TCD have to be combined like an EDR

equation remains a question.

93
 Problems of Expected Data Rate (EDR)

– Even if the link rate is considered in the metric, it does not

consider the fact that multiple rates instead of only the
maximum rate are available in each link.

– Interference range of a given link k is difficult to determine,

so Ik is hard to derive.

– EDR of a routing path cannot take into account the self-

interference,

as the EDR of a bottleneck link is used as the EDR of the

entire routing path.

94
 Airtime Cost Routing Metric
IEEE 802.11s Task Group, Draft, Amendment to Standard for Information
Technology – Telecommunications and Information Exchange Between Systems -
LAN/MAN Specific Requirements – Part 11: Wireless Medium Access Control (MAC)
and physical layer (PHY) specifications: Amendment: ESS Mesh Networking, IEEE
P802.11s/D1.00-2006.).

– To identify an efficient radio-aware path among all the candidate

paths,

– A default routing metric in IEEE 802.11s draft

– Reflects the amount of channel resources consumed for transmitting

a frame over a particular link.

95
 Airtime Cost Routing Metric

– The path which has the smallest sum of airtime cost is the best path.
– The airtime cost Ca for each link is calculated as:

Oca, Op, and Bt depend on the used transmission

technology.
Bt 1 Oca: channel access overhead,
Ca = [Oca + OP + ]
r 1 − e pt Op: protocol overhead,
Bt: number of bits in a test frame.
r and ept : bit rate in Mbit/s and frame error rate for
the test frame size Bt, respectively.

96
 Comparison of Different Routing Metrics

◼ Many routing metrics try to capture link-layer

performance parameters by using a procedure in the
network layer

◼ In fact, these schemes can be enhanced by

performing link-quality measurements directly in the
link layer and then use such measurements in the
network layer

◼ This method implies that the routing metrics should

involve cross-layer interactions

97
 Comparison of Different Routing Metrics

Routing Captured
Layers Performance Advantages Shortcomings
Metrics Parameters

Minimum hop-count is usually not

Hop-count Network Number of hops Simple and low
the performance goal

Packet loss, traffic

load, high overhead in sending probes,
Multiple link metrics
Per-hop RTT Network estimation accuracy depends on
queuing delay, contention captured
traffic load

High overhead in measuring

Packet loss, transmission Multiple link metrics
delay, performance dependent on
Per-hop PPD Network captured, less impacted by
Delay the measurement accuracy, no
traffic load
load balancing
Measurement is not accurate due
Packet loss, to differences between
ETX retransmission, Captured multiple link
broadcast and unicast, no load
Network metrics, relative lower
contention balancing, cannot capture packet
overhead by using broadcast
loss variations, can have
bottleneck link
98
 Comparison of Different Routing Metrics

Routing Captured
Layers Performance Advantages Shortcomings
Metrics Parameters
End-to-end attempts,
ETOP Network, Link position considered in
link Difficulty in deriving the metric
Link routing
retransmission
Same link metrics of
ETT & ETX Improve ETX by considering Same problems of ETX, Not
Network link bandwidth and packet applicable to single-radio multi-
WCETT and also link bandwidth size, channel diversity channel operation
and packet size

99
 Comparison of Different Routing Metrics

Routing Captured
Layers Performance Advantages Shortcomings
Metrics Parameters

BLC MAC handshake, time, Residual capacity of a link is

Network, Bottleneck link of a route does
packet loss rate, hop considered, so load-balancing
Link not consider self-interference
count is performed indirectly
Use contention time of all Have the same problems as those
EDR link metrics as that in
Network interfering links to consider in ETX, hard to find interfering
ETX, contention time
interference links
Overhead in probing, Airtime
Airtime Cost Resource consumed by Captures the impact of the cost captured by probe message
Link dynamic environment to a
a packet on a link may be
link
different from a packet

100
 Remaining Issues

– The measurement or estimation method for a

routing metric may not be accurate

– It may also cause large overhead, especially for a

large scale network.

– Performance comparisons between different routing

metrics need further research, although some work has
been done for a few routing metrics.
101
 Remaining Issues

– The design of many existing routing metrics is still “ad-hoc”,

* i.e., why the proposed routing metric can improve the network
performance is not really justified;

* usually only simulation results are used to prove the effectiveness

of a routing metric.

* the side-effect of such a design is that the effectiveness of a routing metric

may be limited to a certain type of WMNs.

– A routing metric may not be able to capture enough network parameters for a
routing protocol to optimize the network performance.

102
 OVERVIEW OF ROUTING ALGORITHMS

Hop-Count based Routing

Link Level QoS Based Routing
Interference Based Routing (IRMA)
Routing with Load Balancing
Routing Based on Residual Link Capacity
End-to-End QoS Routing
Reliability Aware Routing
Joint Channel Assignment and Routing (Layer 2.5 Routing)

103
 SET 1:
Hop-Count based Routing Algorithms

Light Client Management Routing (LCMR) Protocol

Orthogonal Rendezvous Routing (ORR) Protocol

HEAT Protocol

104
 Light Client Management Routing (LCMR) Protocol
Wehbi B, Mallouli W and Cavalli A, Light client management protocol for
wireless mesh networks.
Proc. 7th Int. Conf. on Mobile Data Management (MDM), 2006

◼ End-to-end routing path from a source to a destination client

consists of
* proactive route among mesh routers and
* reactive routes between clients and mesh routers

◼ To find the best route from one client to another,

hop-count is used as the routing metric.

◼ LCMR does not require routing functionality in clients

◼ Mesh routers supporting clients take care of routing

105
 Light Client Management Routing (LCMR) Protocol

◼ Mesh routers need to maintain two tables:

1. MAC and IP addresses of local clients

2. IP addresses of remote clients as well as the

IP addresses of remote mesh routers associated with
remote clients

106
 Light Client Management Routing (LCMR) Protocol

◼ Based on these two tables,

• When a local client needs to set up a routing path to a remote client,

• its associated mesh router can find out which remote mesh router is
responsible for forwarding traffic to the remote client

• Based on such information, mesh routers can then set up a routing

path between them using proactive routing and hop-count metric.

107
 Disadvantages of Light Client Management Routing (LCMR) Protocol

◼ LCMR has a high overhead of maintaining the two tables on

each mesh routers,

as all clients’ IP addresses need to be collected and stored

at each mesh router.

108
 SET 1:
Hop-Count based Routing Protocols

Light Client Management Routing (LCMR) Protocol

Orthogonal Rendezvous Routing (ORR) Protocol

HEAT Protocol

109
 Orthogonal Rendezvous Routing (ORR) Protocol
Cheng B, Yuksel M and Kalyanaraman S,
Orthogonal rendezvous routing protocol for wireless mesh networks.
Proc. IEEE Int. Conf. on Network Protocols (ICNP), 2006.

◼Each node can define its neighbors’ directions

relative to its local North.

◼Relying on such information, ORR can reduce the

state information for routing, and it does not need
flooding for route construction.

110
 Orthogonal Rendezvous Routing (ORR) Protocol

◼ Compared to geographic routing, ORR does not need exact

location of nodes.

Idea
In 2-D Euclidean space two orthogonal lines can have at
least two intersect points with another group of two
orthogonal lines,

if these two groups of orthogonal lines have different

centers.

111
 Orthogonal Rendezvous Routing (ORR) Protocol

◼ To construct routing paths, a source node sends route

discovery in orthogonal directions,

while a destination node sends route dissemination in

orthogonal directions.

◼ Thus, there is at least one intersect point, called

rendezvous point

where both route discovery and route dissemination

messages are received.
112
 Orthogonal Rendezvous Routing (ORR) Protocol

◼ In this way a routing path is established between the

source and the destination

◼ Also, routing path from the source to the rendezvous point

is a reactive route and

◼ the remaining path to the destination is a proactive route.

113
 Shortcomings of Orthogonal Rendezvous Routing (ORR) Protocol

1. Direction of a node needs to be configured freely

2. Network is not really a 2-D space.

(If a 3-D space is considered, the theory for ORR may not be valid)

3. ORR may not work if the node density is high or topology

change frequently happens

4. Routing path selection procedure is based on hop-count.

(However, other metrics such as link quality can be adopted to enhance the ORR).

114
 SET 1:
Hop-Count based Routing Protocols

Light Client Management Routing (LCMR) Protocol

Orthogonal Rendezvous Routing (ORR) Protocol

HEAT Protocol

115
 HEAT Protocol
Baumann R, Lenders V, Heimlicher S and May M,
HEAT: scalable routing in wireless mesh networks using temperature fields.
Proc. IEEE Int. Symposium on a World of Wireless, Mobile and Multimedia
Networks (WoWMoM), 2007

Anycast routing protocol HEAT considers all nodes

in a WMN as a temperature field
– Gateways have the highest temperature

– Temperature of a non-gateway node will be determined

by hop-count to the gateways and the robustness of a
routing path from this node to gateways.

116
 HEAT Protocol

Once temperatures of all nodes are determined

according to this procedure, the packets from any node
to gateways can simply follow the following method:

The node forwards the packets to its neighbor with the

highest temperature, and this neighbor will repeat the
same process until reaching gateways.

117
 Problems of HEAT Protocol

◼ Totally depends on the assumption that the traffic of

WMNs only needs to be routed between gateways and non-
gateways

◼ For other scenarios, anycast routing is not supported.

◼ Moreover, how to consider other routing metrics in HEAT

remains an open problem

118
 OVERVIEW OF ROUTING ALGORITHMS

Hop-Count based Routing

Link Level QoS Based Routing
Interference Based Routing (IRMA)
Routing with Load Balancing
Routing Based on Residual Link Capacity
End-to-End QoS Routing
Reliability Aware Routing
Joint Channel Assignment and Routing (Layer 2.5 Routing)

119
 Set 2:
Link-level QoS based Routing Algorithms

– Link quality source routing (LQSR) protocol

– Multi-radio LQSR (MR-LQSR) routing protocol
- ExOR Routing Protocol
– AODV-spanning tree (AODV-ST) protocol

120
 Link Quality Source Routing (LQSR) Protocol
Draves R, Padhye J and Zill B,
Comparisons of routing metrics for static multi-hop wireless
networks. Proc. ACM SIGCOMM, 2004.

◼Based on Dynamic Source Routing (DSR)

◼Contains all basic DSR functionalities, such as

* Route Discovery (Route Request and Route Reply messages) and
* Route Maintenance (Route Error messages).

121
 Link Quality Source Routing (LQSR) Protocol

◼ However, LQSR holds two major differences compared to DSR.

– LQSR is implemented as layer 2.5 protocol instead of

as a network layer protocol

– LQSR supports link quality metrics.

122
 Link Quality Source Routing (LQSR) Protocol

◼ Layer 2.5 architecture brings two significant advantages.

– On the one hand, no modification is needed for the higher layer

software,
i.e., LQSR routing protocol is transparent to higher layer software.

– Also no modification is required for link layer software.

123
 Link Quality Source Routing (LQSR) Protocol

◼Performance of LQSR varies based on different

routing metrics and network mobility:

– For stationary nodes in WMNs,

the routing metric ETX, achieves the best performance

124
 Link Quality Source Routing (LQSR) Protocol

For mobile nodes:

Minimum hop count method outperforms the three link
quality metrics, i.e., per-hop RTT, per-hop packet
pair delay, and ETX

REASON:
As the sender moves, the ETX metric cannot quickly
track the changes in the link quality.

125
 Set 2:
Link-level QoS based Routing Protocols

– Link quality source routing (LQSR) protocol

– Multi-radio LQSR (MR-LQSR) routing
protocol
– ExOR Routing Protocol
– AODV-spanning tree (AODV-ST) protocol

126
 Multi-Radio LQSR (MR-LQSR) Routing Protocol
Draves R, Padhye J and Zill B,
“Routing in multi-radio, multi-hop wireless mesh
networks”, Proc. ACM MOBICOM, 2004.

◼ Based on LQSR, and thus, also based on DSR

◼ Major difference from LQSR is WCETT

◼ To make LQSR perform well in a mesh network with

multiple radios per node, WCETT is used as the routing
metric in the routing protocol

127
 Multi-Radio LQSR (MR-LQSR) Routing Protocol

◼WCETT takes into account both link quality metric

and the minimum hop-count

◼Achieves good tradeoff between delay and

throughput

because it considers channels with good quality and

channel diversity in the same routing protocol
128
 Multi-Radio LQSR (MR-LQSR) Routing Protocol

◼MR-LQSR assumes nodes are stationary

◼This is true for mesh routers, but obviously not

applicable to mesh clients

◼Performance of MR-LQSR can also be degraded by

the mobility of nodes, i.e., mesh clients.

129
 Multi-Radio LQSR (MR-LQSR) Routing Protocol

◼In WMNs, multi-channel operation over a single

radio is another alternative to increase the network
capacity.

◼But MR-LQSR is not applicable because WCETT is

limited to multi-radio mode.

130
 Set 2:
Link-level QoS based Routing Protocols

– Link quality source routing (LQSR) protocol

– Multi-radio LQSR (MR-LQSR) routing protocol
–ExOR Routing Protocol
– AODV-spanning tree (AODV-ST) protocol

131
 ExOR Routing Protocol
Biswas S and Morris R, “ExOR: opportunistic multihop routing for
multi-Hop wireless networks,” in Proc. ACM SIGCOMM, 2005.

* Proposed to improve throughput based on

cooperative broadcasting packets from source to
destination without explicitly setting up a routing
path

132
 Source’s behavior

◼ Collects enough packets of the same destination to form a

batch
– ExOR operates on batches of packets for efficiency

◼ Selects a set of nodes to be candidate forwarders, and

includes the prioritized list in the overhead of every packet

133
 ExOR Routing Protocol

* This priority of a forwarding node is determined by

the cost to the destination,

which is evaluated by the accumulative ETX to the

destination node.

134
 ExOR Routing Protocol

◼Set of nodes that are selected for forwarding

packets are determined based on the packet loss
ratio between the source and these nodes.

◼Although many nodes can receive packets from the

source node, only a subset of nodes are selected
as forwarding nodes
in order to reduce the overhead.
135
 Forwarders’ Behavior

◼How can a node know whether it is one of the

forwarders or not?
◼Check the forwarder list in the overhead of the
received packet
– If the node finds itself in the list, buffer the packet
and keep state of this batch
– If no, discard the packet

136
 Forwarder’s Behavior

▪ Highest priority forwarding node sends its own batch of

packets following the same procedure as done by the
source node.

▪ This process is repeated until 90% of packets in each

batch are received by the destination node.

▪ The remaining 10% of nodes will rely on traditional minimum

hop-count routing to deliver.
137
 Forwarders’ Behavior

◼How can a node know whether the packet it receives

has also been received by a node with higher priority
or not?

◼ExOR designs a “batch map” to record, for every

packet in the batch, the highest-priority node known
to have received that packet.

138
 Advantages of ExOR Routing Protocol

◼ Most of packets are delivered without setting up routing

paths, which is similar to anycast routing.

◼ Moreover, ExOR can improve throughput for two reasons.

– It tries to use the best link to deliver packets through
cooperation of forwarding nodes.

– Progress of packet forwarding can be continued even if some nodes

on a traditional path experiences bad link quality or out of order.

139
 Set 2:
Link-level QoS based Routing Protocols

– Link quality source routing (LQSR) protocol

– Multi-radio LQSR (MR-LQSR) routing protocol
– ExOR Routing Protocol
– AODV-spanning tree (AODV-ST) protocol

140
 AODV-Spanning Tree (AODV-ST) Protocol
Ramachandran K, Buddhikot MM, Chandranmenon G, Miller S,
Almeroth K and Belding-Royer E,
On the design and implementation of infrastructure mesh networks.
Proc. IEEE WIMESH, 2005.

◼ AODV ST is designed for multi-radio WMNs

◼ ADOV-ST performs hybrid routing,

i.e., for traffic inside the mesh network AODV is used

where the spanning-tree based routing is used for traffic

to/from gateways.

◼ ETT is used as the routing metric

141
 Drawbacks of AODV-Spanning Tree (AODV-ST) Protocol

◼AODV may not be efficient for intra-mesh traffic

◼The WCETT proposed for multi-radio WMNs is not

applicable,
because AODV is a distance vector routing protocol

142
 OVERVIEW OF ROUTING ALGORITHMS

Hop-Count based Routing

Link Level QoS Based Routing
Interference Based Routing (IRMA)
Routing with Load Balancing
Routing Based on Residual Link Capacity
End-to-End QoS Routing
Reliability Aware Routing
Joint Channel Assignment and Routing (Layer 2.5 Routing)

143
 Set 3:
Interference Based Routing: IRMA
Wu Z, Ganu S and Raychaudhuri D,
IRMA: integrated routing and MAC scheduling in multihop
wireless mesh networks. Proc. IEEE WiMesh, 2006.

◼ TDMA instead of CSMA/CA as MAC

◼ Based on the TDMA, an integrated routing and MAC

scheduling algorithm (IRMA) is derived to find a routing
path for each traffic flow

◼ then determine slot allocation on each link considering

* BW allocation information
* Link status, and
* Topology information in the network.
144
 Interference based Routing: IRMA

◼ A centralized scheme

◼ Relies on an existing solution to collect the node, link, and

topology related information of the entire network and get
traffic specifications of traffic flows.

◼ Such signaling can be done in a global control plane (GCP)

that can be implemented in a separate dedicated channel
or a dedicated time slot

145
 Interference based Routing: IRMA

Two routing mechanisms defined in IRMA:

– link scheduling with minimum-hop routing

– link scheduling with bandwidth-aware routing (preferred)

146
 Interference Based Routing: IRMA
Link Scheduling with Minimum-Hop Routing

◼ A routing path is selected by shortest path using minimum

hop-count.

◼ Then time slots along this path are determined by the

centralized algorithm by considering the latest flow
information in the network.

→may result in congested paths or links as the minimum hop-

count routing does not consider the available BW.
147
 Interference Based Routing: IRMA
Link Scheduling with bandwidth-aware Routing

◼ Available BW on each link is factored when a routing path is selected.

◼ Based on the selection, time slots are then determined for each link
on the path.

◼ Thus, such a scheme can not only avoid contentions in traffic flows
but can also avoid bottleneck or congested links.

148
 Shortcomings of Interference Based Routing: IRMA

1. Not scalable with the network size (a centralized scheme)

2. Assumes an efficient scheme to collect all control information,

which is a challenging issue for all routing protocols.

3. MAC layer is assumed to have TDMA operation, which is not the

case for many WMNs.

149
 OVERVIEW OF ROUTING ALGORITHMS

Hop-Count based Routing

Link Level QoS Based Routing
Interference Based Routing (IRMA)
Routing with Load Balancing
Routing Based on Residual Link Capacity
End-to-End QoS Routing
Reliability Aware Routing
Scalable Routing
Joint Channel Assignment and Routing (Layer 2.5 Routing)

150
 Set 4:
Routing with Load Balancing
Song W and Fang X ,
”Routing with congestion control and load balancing in wireless mesh networks,’
Proc. Int. Conference on ITS Telecommunications, 2006

◼ Routing is directly determined by considering network congestion

◼ Given a source and its destination, a routing path is determined by

using the route with the least congestion

◼ If more than one paths have the same number of congested nodes,
the route with minimum hop-count is selected

◼ Congestion state of a link is determined by the number of

retransmissions of RTS and ACK packets.

◼ If the congestion exceeds a threshold, the congestion weight on this

link increases
151
 Routing with Load Balancing: CAR
Liu T and Liao W,
Capacity-aware routing in multi-channel multi-rate wireless mesh
networks. Proc. IEEE ICC, 2006.

◼ A capacity-aware routing (CAR) protocol is proposed to balance load

among links and channels in a multi-radio WMN

◼ CAR assumes channel assignment on radios lasts long time and can
be static.

◼ With static channels on each radio in the network, CAR determines

the BLC in a reactive manner for each traffic flow.

152
 Routing with Load Balancing: CAR

◼Transmissions start on a routing path that is

determined according to the best BLC metric.

◼However, this path can be switched to a new one if

the source finds the new path has a higher BLC
value.

153
 Routing with Load Balancing: CAR

◼Due to the bottlenecked link capacity in routing,

CAR improves throughput and delay performance.

◼ CAR may not achieve optimal performance because

the path selection on different flows affect each
other but is not coordinated among such flows.

154
 OVERVIEW OF ROUTING ALGORITHMS

Hop-Count based Routing

Link Level QoS Based Routing
Interference Based Routing (IRMA)
Routing with Load Balancing
Routing Based on Residual Link Capacity
End-to-End QoS Routing
Reliability Aware Routing
Scalable Routing
Joint Channel Assignment and Routing (Layer 2.5 Routing)

155
 Set 5:
Routing Based on Residual Link Capacity
Raniwala A, Gopalan K and Chiueh T ,”Centralized channel assignment and routing
algorithms formulti-channel wireless mesh networks,”
ACM Mobile Computing and Communications Review, 2005.

◼ An alternative scheme to consider link capacity in routing is

to get the information of residual link capacity

◼ A protocol called Hyacinth is developed to perform routing

and channel assignment for a multi-channel WMN.

156
 Routing Based on Residual Link Capacity

◼ Hyacinth considers traffic from/to gateways in WMNs and

thus uses tree-based routing for such traffic.

◼ Each node advertises its costs of its path from/to the

gateway

◼ Based on such information, a neighbor that finds a lower

value in the cost will leave its old parent node and selects
the new node as the new parent node.

157
 Routing Based on Residual Link Capacity

◼ With such a procedure, all nodes in the network build up

routing paths to the gateway like a spanning tree

◼ To reflect the cost of a path,

residual capacity of a link → the routing metric.

◼ Links are selected which have the largest available capacity

158
 OVERVIEW OF ROUTING ALGORITHMS

Hop-Count based Routing

Link Level QoS Based Routing
Interference Based Routing (IRMA)
Routing with Load Balancing
Routing Based on Residual Link Capacity
End-to-End QoS Routing
Reliability Aware Routing
Scalable Routing
Joint Channel Assignment and Routing (Layer 2.5 Routing)

159
 Set 6:
End-to-End QoS Routing

◼ Quality Aware Routing Protocol

◼ Ring Mesh Routing Protocol

160
 Quality Aware Routing Protocol
Koksal CE and Balakrishnan H,
”Quality-aware routing metrics for time-varying wireless
mesh networks,”
IEEE Journal on Selected Areas in Communications, 2006.

◼ End-to-end QoS can be considered in a routing protocol by

ensuring end-to-end packet loss rate, delay, or bandwidth.

◼ End-to-end packet loss rate is ensured

where both ETX and ENT are used as routing metrics

◼ ENT is used to determine what routing paths can be used

161
 Quality Aware Routing Protocol

− Allowed packet loss rate in a link is given

− Based on this packet loss threshold and the measurement of the

link, a positive number δ used by ENT is derived.

− With δ and an existing probe scheme, ENT of each link is determined.

162
 Quality Aware Routing Protocol

− ENT is then compared with the maximum number of

transmissions of a packet before it is discarded at the
link layer

− If ENT is larger than the link-layer value, the routing

cost of the link will become ∞

− Otherwise, ETX is used for the routing cost of the link

163
 Quality Aware Routing Protocol

− As a result, all links that do not satisfy packet loss

requirement will be excluded from routing paths

− After this step, routing path selection is performed by just

choosing a path with smallest routing cost

◼ One critical issue:

How to determine the allowed packet loss rate of a link
given the threshold of end-to-end packet loss requirement?
164
 Set 6:
End-to-End QoS Routing

◼ Quality Aware Routing Protocol

◼ Ring Mesh Routing Protocol

165
 RingMesh Routing Protocol
Lin D, Moh T and Moh M ,”A delay-bounded multi-channel routing
protocol for wireless mesh networks using multiple token rings:
extended summary,”
Proc. 31st IEEE Conference on Local Computer Networks (LCN), 2006.

◼ End-to-end delay

◼ RingMesh is developed based on a token ring

protocol proposed for wireless LANs

166
 RingMesh Routing Protocol

◼ Multiple token rings are created and organized from the

gateway to all other nodes like a spanning tree scheme

◼ Different channels are used in neighboring rings

◼ First ring containing the gateway is called a root ring

◼ Next ring connected to the root ring is a child ring.

167
 RingMesh Routing Protocol

◼ These two rings share a common node which is called a

pseudo gateway

◼ Following this process, other child rings are connected

together all the way to the root ring

◼ For a node in the network, which ring it can join depends on

the delay from it to the gateway:

◼ the node joins a ring that can satisfy the end-to-end delay
requirement.
168
 Shortcomings of RingMesh Routing Protocol

− How to form multiple rings to support multiple gateways to improve

the delay performance is not addressed?

− It is also unknown what happens if no ring can be joined by a node

− No mechanism is also available to determine the delay from a node

to the gateway, which is not a trivial task

− Thus, end-to-end delay aware routing is still a challenging

research topic

169
 OVERVIEW OF ROUTING ALGORITHMS

Hop-Count based Routing

Link Level QoS Based Routing
Interference Based Routing (IRMA)
Routing with Load Balancing
Routing Based on Residual Link Capacity
End-to-End QoS Routing
Reliability Aware Routing
Scalable Routing
Joint Channel Assignment and Routing (Layer 2.5 Routing)

170
 Set 7:
Resilient Opportunistic Mesh Routing
(ROMER) Protocol
Yuan Y, Yang H,Wong SHY, Lu S and Arbaugh W,”ROMER: resilient
opportunistic mesh routing for wireless mesh networks,”
Proc. IEEE WIMESH, 2005

◼ ROMER creates forwarding mesh on the fly for each packet

◼ ROMER assumes that there is an existing scheme that can

find the minimum cost from each mesh router to the gateway

◼ Then the credit is determined while a packet is forwarded on

the fly

171
 Resilient Opportunistic Mesh
Routing (ROMER) protocol

◼ When a packet is to be delivered from a mesh router, e.g., Node S,

to the gateway, the source mesh router needs to set a credit cost.

◼ If the minimum cost from S to the gateway is Cmin,S and the credit
cost is Ccredit,S,
then S has a budget cost of Cmin,S + Ccredit,S to the gateway.

◼ When the packet is sent to the next mesh router, e.g.,

node A, the budget is reduced by the cost of the
traversed link clinkSA

172
 Resilient Opportunistic Mesh
Routing (ROMER) protocol

◼ At mesh router A, the needed credit is computed according

to the requirement of
Ccredit,A + Cmin,A + ClinkSA <= Cmin,S + Ccredit,S,

i.e., the remaining credit at mesh router A is

Ccredit,A = Cmin,S + Ccredit,S − Cmin,A − ClinkSA.

◼ If the ratio of the remaining credit over the initial credit

Ccredit,S is less than a threshold, e.g., (Cmin,A / Cmin,S)2 ,
then the packet at mesh router A shall be discarded;
173
 Resilient Opportunistic Mesh
Routing (ROMER) protocol

◼ Otherwise, mesh router A forwards the packet according to

a randomized opportunistic forwarding scheme.

◼ The above process is repeated until the packet is delivered

to the gateway

174
 Resilient Opportunistic Mesh
Routing (ROMER) protocol

◼ Finally, when multiple intermediate routers receive the

same packet from a mesh router and

all have enough credit to forward the packet,

they need to follow a randomized opportunistic forwarding

scheme to forward packets.

175
 Resilient Opportunistic Mesh
Routing (ROMER) protocol

◼Probability that each intermediate router can

forward a packet depends on the quality of the link
to the parent router

176
 Resilient Opportunistic Mesh
Routing (ROMER) protocol
Intermediate router with the best link quality
forwards the packet with probability 1,

while other intermediate routers forward the packet

with a probability of (Rl / Rmax),

where Rl is the current rate of the considered link

and Rmax is the current rate at the best link.

177
 Drawbacks of Resilient Opportunistic
Mesh Routing (ROMER) Protocol

◼ ROMER has to rely on an existing scheme to find out the

minimum cost from each mesh router to the gateway

◼ What type of cost is the best for ROMER and how to

dynamically update the cost to best serve ROMER remain
open questions !

178
 OVERVIEW OF ROUTING ALGORITHMS

Hop-Count based Routing

Link Level QoS Based Routing
Interference Based Routing (IRMA)
Routing with Load Balancing
Routing Based on Residual Link Capacity
End-to-End QoS Routing
Reliability Aware Routing
Joint Channel Assignment and Routing (Layer 2.5
Routing)

179
 Multichannel Protocols

◼ Multi-channel operation is widely adopted in WMNs to

improve network capacity

◼ Single-channel routing protocols may be run in each of the

channels of the WMN
– Easy design but not optimal, and does not guarantee availability of spectrum in the
routes

◼ Multi-channel routing protocols are better suited

180
 Multichannel Protocols

◼ Two types of multi-channel routing protocols

–Type 1: Consider the impact of multi-channel operation such as link

quality, interference, packet loss, bandwidth

–Type 2: Conduct close routing/MAC cross-layer design such as joint

channel allocation and routing

181
 Multichannel Protocols

◼ Most existing protocols are of type 1

– As an example, in MQ-LSR ignores the close relationship between

traffic distribution and channel allocation by assuming different
radios are assigned non-overlapping channels

– It also incorrectly assumes that the channel assignment changes

relatively infrequent, leading to the necessity of joint channel-route
assignment

182
 Joint Channel Assignment and Routing
Alicherry M, Bhatia R and Li L, “Joint channel assignment and routing
for throughput optimization in multi-radio wireless mesh networks,”
in Proc. ACM MobiCom, 2005

◼ For infrastructure WMNs (IWMNs)

Assumption:
Aggregated traffic load at mesh routers and channels assigned to
each router is not changing frequently

◼ Channels assigned to radios on a node are determined together with

routing paths
with an objective to obtain interference-free link schedule and
achieve maximum throughput.

183
 Joint Channel Assignment and Routing
Tang J, Xue G and Zhang W, “Interference-aware topology control
and QoS routing in multichannel wireless mesh networks,”
ACM MobiHoc, 2005.

◼ Assumes the channel assignment can be static in WMNs

◼ Goal: mathematical formulation of the joint design between

channel assignment and routing

◼ However, no actual protocol is proposed

184
 Distributed Joint Channel and Routing Protocol
Avallone S and Akyildiz, IF and Ventre G, “A channel and rate assignment
algorithm and a layer-2.5 forwarding paradigm for multi-radio wireless
mesh neetworks,” IEEE/ACM Transactions on Networking, 2009

◼It takes into account the number of flows that are

possible to route on each link

◼Amount of flows is obtained from a solution to the

joint channel assignment and routing problem

185
 Distributed Joint Channel and Routing Protocol

◼ Objective
Enable every router to utilize each of its links in proportion
to their assigned flow rates

◼ Routing protocol only requires a partial knowledge of the

network topology and does not make use of a destination-
based routing table

◼ Hence, the name Layer-2.5 (L2.5) given to the routing

protocol.

186
 Layer-2.5 Routing Algorithm

◼ Each mesh router is configured with the set of pre-

computed flow rates associated with its links
◼ Packets are forwarded using such information (rather
than routed using routing tables)
– Layer-2 information are used, hence the name
◼ Each mesh router attempts to keep the utilization of
the outgoing links proportional to their pre-computed
flow rates

187
 Channel Assignment & Routing

* An approximate solution:
Determine pre-computed flow rates
A pre-computed flow rate is determined for every link based on the
given optimization objective

Determine the channel assignment

Channels are assigned to radios in the attempt to make such
pre-computed flow rates schedulable

Adjust the pre-computed flow rates

The pre-computed flow rates may be adjusted in order to obtain a set
of schedulable flow rates given the computed channel assignment

188
 Open Research Issues

◼ Performance Benchmark
– Large number of routing metrics and routing protocols available for
WMNs.

– Considering routing protocols for other multi-hop wireless networks,

the number is much bigger

– Confusion about which routing metric and what type of routing

protocols can provide the best performance => benchmark to
investigate and compare different routing metrics and protocols.

189
 Open Research Issues
– It is expected to include theoretical analysis of performance
bound, practical consideration of protocol design, and
performance evaluation either through simulations or testbeds.
⚫ In [1], a comparative study is carried out for different routing strategies for
WMNs.
⚫ Some design guidelines are provided in [2] for multihop wireless networks.
⚫ However, such work is still far from providing a benchmark of selecting
routing metrics and protocols.

[1] Wellons J, Dai L, Xue Y and Cui Y, “Predictive or oblivious: a comparative study of Routing strategies
for wireless mesh networks under uncertain demand," Proc. IEEE SECON, 2008
[2] Yang Y and Wang J, “Design guidelines for routing metrics in multihop wireless networks,”
Proc. IEEE INFOCOM, 2008.

190
 Open Research Issues

◼New routing metrics

– How to integrate multiple routing metrics into the same
routing protocol is another challenging issue.

191
 Open Research Issues

◼ Scalable routing
– This is a critical requirement by WMNs, achieved by few routing
protocols so far.
– Hierarchical routing protocols can only partially solve this problem
due to their complexity and difficulty of management.
– Geographic routing needs GPS or similar  cost, complexity.
Additional traffic by inquiry of destination position.
– Scalability is also related to MAC protocols. Thus, an eventual
scalable routing protocol must be closely integrated with the MAC
protocol.

192
 Open Research Issues
◼ Network coding and routing
– Can potentially improve the performance of WMNs
E.g., research to apply network coding to WMNs [1], [2],
Benefits of network coding to a multichannel WMN in [2]
– However, as network coding is still in an early phase of being
applicable to a practical networking protocol, integrating network
coding with routing is still a new and challenging research direction

[1] Omiwade O, Zheng R and Hua C, “Practical localized network coding in wireless mesh networks,”
Proc. of IEEE SECON, 2008

[2] Zhang X and Li B, “On the benefits of network coding in multi-channel wireless networks,”
Proc. of IEEE SECON, 2008

193