SRL: A Bidirectional Abstraction for
Unidirectional Ad Hoc Networks.
Venugopalan Ramasubramanian
Ranveer Chandra
Daniel Mosse
Introduction
 Links in an ad hoc network could be
unidirectional.
 Many Ad hoc network routing protocols are
not designed to handle unidirectional links
(TORA).
 Some handle unidirectional links but are very
inefficient (DSR).
Noise: source of one-way link.
C
E
A
D
B
 Transient unidirectional links.

Go away when noise subsides or nodes move.
Asymmetry in Transmit Power
C
C
B
A
B
A
 Topology Control Schemes: Sensor Network
 Heterogeneity of hardware: Home Network
Problems due to one-way links.
 Collision avoidance (RTS/CTS) scheme is
impaired

Even across bidirectional links!
A
B
C
RTS
CTS
MSG
CTS
X
MSG
MSG
Problems due to one-way links
 Collision avoidance (RTS/CTS) scheme is
impaired

Even across bidirectional links.
 Unreliable transmissions through one-way
link.

May need multi-hop Acks at Data Link Layer.
 Link outage can be discovered only at
downstream nodes.
Problems for Routing Protocols
 Route discovery mechanism.


Cannot reply using inverse path of route request.
Need to identify unidirectional links. (AODV)
 Route Maintenance.

Need explicit neighbor discovery mechanism.
 Connectivity of the network.

Gets worse (partitions!) if only bidirectional
links are used.
Bidirectional Connectivity(%)
Average Bidirectional Connectivity
110
100
90
80
70
60
50
40
P5
P 10
P 15
P 20
P 25
Probablity of one-way link (%)
P 30
P 35
Distribution of Bidirectional Connectivity.
200 random topologies. Probablity of one-way link = 0.25
50
45
35
30
25
20
15
10
5
Connectivity (%)
10
0
95
90
85
80
75
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0
0
Distribution
40
Reverse route for one-way link
 Let A  C be a one-way link.
 C  B  A is a 2-hop reverse route.
C
B
A
Connectivity with reverse routes.
100%
95%
90%
Rts 5
85%
Rts 4
80%
Rts 3
Rts 2
75%
Rts 1
70%
65%
60%
P5
P 10
P 15
P 20
P 25
P 30
P 35
One-way links with reverse routes.
100%
90%
80%
len 4 %
len 3 %
70%
len 2 %
60%
len 1 %
50%
40%
P5
P 10
P 15
P 20
P 25
P 30
Probablity of unidirectional links (%)
P 35
Average Reverse Route Length
reverse route length (hops)
1.6
1.5
1.4
1.3
1.2
1.1
1
P5
P 10
P 15
P 20
P 25
P 30
Probablity of unidirectional link (%)
P 35
Observations from analysis.
 Topologies generated with asymmetric
transmit power also produce similar graphs.
 The connectivity follows a long tail
distribution.
 Reverse routes are short (2 or 3 hops) for
most one-way links.
SRL: Sub Routing Layer
 Short reverse routes for one-way links


Improve connectivity substantially.
Also decrease route lengths.
 SRL discovers and maintains reverse routes for
one-way links.
 It provides a bidirectional abstraction to the routing
protocols.
 Provides services such as reliable transmission and
link breakage detection.
Internals of SRL
 Reverse Distributed Belmanford Algorithm

Distance vector based technique.
 Each node maintains:


Shortest path from other nodes in its locality.
Periodically neighbor-casts this information.
 Locality of node A:


Set of nodes that can reach A in r hops.
r: is the radius of locality.
Reverse Distributed Belmanford Algorithm.
Reverse Route: C  B  A
A; 1; C
A; 2; C
C; 1; B
C
C; 2; B
B; 1; A
B
A
Update Message Format: Source; #hops; First Hop
RDBA contd.
 Periodic update messages are neighbor-cast:
Source ID : Hop Count : First Hop
 Sources restricted to locality of radius r.


r: called SRL radius is small (2 – 3).
Scalable to large networks.
 No counting to infinity problem.

Ignore distances bigger than r.
 No Route-loops.

Use first hop information to check for loops.
SRL: Periodic Updates
 Incremental Updates



Most recent changes in hop count or first hop.
Sent periodically at same rate as hello messages.
Replaces hello messages.
 Complete Updates



Contains entire data for locality.
Sent with much lower frequency.
Random distribution to avoid co-ordination.
 Hello Packets

Sent when no incremental updates need to be sent.
Optimization 1: Dynamic SRL
 The SRL radius of each node could be different.
 Each node increases radius until it can find reverse
routes.
 Radius decreases if reverse routes are shorter than
the radius.
 Decreases the number of updates that is neighborcast: lower overhead.
Optimization 2: On-demand DSRL
 Routing protocol requests DSRL to find
reverse routes for certain one-way links.
 Reverse routes maintained only for the
chosen one-way links.
 Routing strategy that uses one-way links only
when route discovery along bidirectional
links fail.
Services provided by SRL
 Identification of one-way links (radius = 1):

Routing protocols can avoid them.
 Reverse route forwarding:


Routing protocol uses reverse routes to send route replies and route
errors.
Not good for data packets.
 Link breakage detection:

Several protocols rely on lower layers to do this.
 Reliable Transmission across unidirectional links:

Multi-hop Acks can be used if required by the protocol.
Simulation: AODV over SRL
 AODV is adapted on top of SRL.


Use reverse routes for RREPs and RERRs.
Uses SRL’s link break discovery service.
 Compared with traditional AODV.



Routes only along bidirectional links.
Uses black-list to identify unidirectional links.
Runs on top of IEEE 802.11
Simulation Setup
 80 nodes in 1300m x 1300m area.
 220m nominal radio range (WaveLan).
 360s total simulation time.





300s of data origination.
20 random src-dest pairs for each run.
50 random topology for each experiment.
Packet Size: random between 64B – 1024B.
Average data rate: 1 packet per sec.
Static Experiments:
Packet Delivery.
# Packets Delivered.
200000
150000
RADIUS 1
100000
RADIUS 2
RADIUS 3
AODV
50000
D0
D 60
D 140
D 220
D 280
Average Diversity in Range (m)
D 320
Static Experiments: Average
Route Length.
Average #hops per packet.
6
5.5
5
4.5
RADIUS 1
4
3.5
RADIUS 2
RADIUS 3
AODV
3
D0
D 60
D 140
D 220
D 280
Average Diversity in Range (m)
D 320
Mobility Experiments:
Packets Originated
transmission range between 80m and 360m
#Packets Originated
100
80
60
RADIUS 1
40
RADIUS 2
RADIUS 3
20
AODV
0
P0
P 60
P 120 P 180 P 240
pause time (sec)
P 300
P 360
Mobility Experiments:
Packet Delivery.
#packets delivered
transmission range between 80m and 360m.
100
80
60
RADIUS 1
40
RADIUS 2
20
RADIUS 3
AODV
0
P0
P 60
P 120 P 180 P 240
pause time (sec)
P 300
P 360
SRL Overhead: Average Length
of Update Packets.
Transmission Range between 80m and 360m
Packet Length (Bytes)
250
RADIUS 0
200
RADIUS 1
RADIUS 2
150
100
50
0
P0
P 60
P 120 P 180 P 240
Pause time (sec)
P 300
P 360
Conclusions
 SRL increases the packet delivery of AODV by
30%.
 The overhead generated by SRL is not very
significant and can be further reduced.
 The effect of optimizations need to be studied.
 RTS/CTS implementation with SRL would be
interesting!