Analysis of Inconsistent Routing Components in Reactive Routing Protocols Habib-ur Rehman , Lars Wolf Institut für Betriebssysteme und Rechnerverbund Technische Universität Braunschweig WMAN 2009, 5 th March, Kassel
Analysis of Inconsistent Routing Components in Reactive Routing
Protocols
Habib-ur Rehman, Lars Wolf
Institut für Betriebssysteme und Rechnerverbund
Technische Universität Braunschweig
WMAN 2009, 5th March, Kassel
2/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Outline
IntroductionHow to Improve Reactive Routing?
The Problem: Use of Prior-to-demand Collected Routing Data??
Analysis of AODVObjectives and Nature of Analysis
AODV-TTL
AODV-RS
Simulation Setup/Results
Conclusions
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
3/17IBR, TU BraunschweigInconsistent Reactive Routing Components
How to Improve Reactive Routing?
• Reactive Routing• On-demand operations
• High Response Time• Connection set up/recovery
• Typical Approach• Use prior-to-demand collected routing data
• Share more-than-demanded routing data• route request/reply packets carry additional data
• Collect more-than-required routing data• overhear the routing packets for others
• Use in route interruptions or subsequent route discoveries
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
4/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Use of Prior-to-demand Collected Routing Data
• Examples• DSR maintains alternate routes by overhearing routing packets
• AODV uses previously known hop-count in new route discoveries
• Overhearing: a common practice among multiple path protocols
• For example: AOMDV, AODV-BR
• An Inconsistent Approach• No proactive mechanism to refresh stored routing data
• Due to ever changing topology future and fortune of such acts• Totally dependent on network and topology conditions
• Unpredictable and volatile behavior/effects/benefits
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
5/17IBR, TU BraunschweigInconsistent Reactive Routing Components
This Paper
• Analyze: use of prior-to-demand collected routing data• Understand the effect on
• Protocol operations
• Protocol/Network performance
• Approach• Analyze the deviation in the behavior of a reactive routing
protocol after• Increasing the use of previously collected routing data
• Decreasing the use of previously collected routing data
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
6/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Analysis
• Standard AODV vs. two modified versions• AODV-TTL
• less dependent on previously collected routing data
• more reactive
• AODV-RS• shares more routing data for subsequent use
• subsequent actions: less reactive
• Compared performance metrics• MAC overhead
• Routing overhead
• Data packet delivery ratio
• Route discovery time
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
7/17IBR, TU BraunschweigInconsistent Reactive Routing Components
AODV-TTL
• Expanding ring search during the route discovery• TTL field determines how many hops a RREQ will travel
• In AODV: in case of an existing entry• TTL = last known hop count + TTL_INCREMENT > TTL_START
• In AODV-TTL• TTL = TTL_START
• Route recovery or route discoveries: completely on-demand
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
8/17IBR, TU BraunschweigInconsistent Reactive Routing Components
AODV-RS
• Routing messages carry the information on two nodes only• The originator and the previous hop
• Route Sharing• Include all the nodes along the path into a RREQ/RREP message
• In AODV-RS
• every intermediate node appends its previous hop
• shares ample amount of prior-to-demand routing data
• effect the subsequent actions
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
9/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Simulations
• OPNET Modeler• manet_station node model
• Random way point mobility
• Simulation scenarios• Varying network size and data streams
• Varying mobility parameters
Simulation Scenarios
Nodes AreaData
StreamsActiveNodes
5 8
20 20
20 30
80 85100
2000 mX
500 m
25800 m
X800 m
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
1, 2, 5, 10, 2010Packet Rate
41, 2, 5, 10, 250Node Speed
410, 30, 60, 300,900, 1800
Pause Time
Data Packet Rate(packets/second)
Node Speed(m/sec.)
Pause Time(seconds)
Variationof
Simulation settings forPause Time, Node Speed and Packet Rate
10/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Results: MAC Overhead
• AODV-RS• 2-20 % higher
• AODV-TTL• 1-11 % less
MAC Ov erhead (25 nodes 5 stream s)
395.8248 388.3791 384.4953
0
80
160
240
320
400
pa
cket
s (x
10
00
)
A ODV -RS A ODV A ODV -TTL
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
MAC Overhead (100 nodes 80 stream s)
2481.7781
2052.84171847.5576
0
500
1000
1500
2000
2500
3000
pa
cket
s (x
10
00
)
A ODV -RS A ODV A ODV -TTL
11/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Results: Routing Overhead
• AODV-RS• 2-20 % higher
• AODV-TTL• 1-11 % less
• Quite similar to MAC overhead• In reactive routing protocols,
Routing traffic dictates the overhead
Routing Ov erhead (25 nodes 5 stream s)
17.8699 17.5296 17.3544
0
4
8
12
16
20
pa
cket
s (x
10
00
)
A ODV -RS A ODV A ODV -TTL
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
Rou t ing Ov erh ea d (100 n odes 80 st ream s)
559.4331
462.4367416.193
0
125
250
375
500
625
pa
cket
s (x
10
00
)
A ODV -RS A ODV A ODV -TTL
12/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Results: Overhead
• Why the packet overhead is high in AODV-RS?• Higher initial value of TTL
• Less controlled flooding
• Higher contribution of RREP messages
• More nodes are able to respond during route discovery
Initial value of the TTL field
AODV-RS AODV
25 nodes 5 stream 1.69 1.21
25 nodes 20 streams 2.27 1.52
100 nodes 20 streams 3.03 1.81
100 nodes 80 streams 4.77 2.56
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
19.2925.34100 nodes 80 streams
18.5921.73100 nodes 20 streams
17.1018.3225 nodes 20 streams
13.5313.6625 nodes 5 stream
RREP
77.8472.69100 nodes 80 streams
78.1075.89100 nodes 20 streams
80.5377.6925 nodes 20 streams
82.4281.7325 nodes 5 stream
RREQ
AODVAODV-RS
Percentage of RREQ and RREP packets
13/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Results: Packet Delivery
• Data Packet Delivery Ratio• AODV-RS
• 1-10 % less
• AODV-TTL
• 1-8 % higher
• Higher overhead
• causes more saturation
• results in less throughput
Pa cket Deliv ery Ra t io (25 n odes 5 st rea m s)
0.9609 0.9706 0.9803
0.6
0.7
0.8
0.9
1
A ODV -RS A ODV A ODV -TTL
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
Pa cket Deliv ery Ra t io (100 n odes 80 st ream s)
0.6834
0.7594
0.8277
0.6
0.7
0.8
0.9
1
A ODV -RS A ODV A ODV -TTL
14/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Results: Route Discovery
• Route Discovery Time• Inconclusive
• 802.11 is a contention-based MAC
• AODV-RS• 3 % less in (25 nodes 5 streams)
scenario
• 2-6 % higher in others
• RREP requires RTS/CTS exchange
• AODV-TTL• 1 % less in (100 nodes 20 streams)
scenario
• 0.5-3 % higher in others
• Requires more expansion steps of ring search
Rou t e Discov ery T im e (25 n odes 5 st rea m s)
0.3809
0.3929 0.3947
0.32
0.34
0.36
0.38
0.4
0.42
seco
nd
s
A ODV -RS A ODV A ODV -TTL
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
Rou t e Discov ery T im e (100 n odes 20 st rea m s)
1.1587
1.09531.0831
1
1.04
1.08
1.12
1.16
1.2
seco
nd
s
A ODV -RS A ODV A ODV -TTL
15/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Conclusions
• More prior-to-demand routing data present in the network• Less RREQs but more RREPs
• AODV loses the benefit of expanding ring search
• suffers due to higher TTL
• More overhead
• AODV-RS > AODV > AODV-TTL
• Less packet delivery ratio
• Mainly due to higher overhead, contention
• Route discovery time
• unpredictable in contention based scenarios
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
16/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Conclusions
• Expanding ring search without exceptions• Less overhead
• Higher route discovery time
• Sharing more routing data: Not a good approach• Higher overhead
• Collecting more routing data might work in some cases
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
Thank you very much for your attention
Questions/Comments/Suggestions
Introduction Analysis AODV-TTL AODV-RS Simulations Results ExtrasConclusions
18/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Why AODV?
• Popular and well reputed
• A very simple protocol
• Based on fundamental reactive principles• Route discovery
• purely reactive
• except the TTL adjustment
• expanding ring search: a good approach to control flooding
• Presence of prior-to-demand routing data• Works the same in most of the reactive protocols
AODV TTL Simulation Parameters Results Main
19/17IBR, TU BraunschweigInconsistent Reactive Routing Components
AODV Routing Protocol
• Route Discovery• Floods RREQ, unicast RREP
• Expanding ring search approach
• start TTL with TTL_START
• step by TTL_INCREMENT on every failed attempt
• until reaches NET_DIAMETER
• in case of an existing entry, start TTL withHOP_COUNT+TTL_INCREMENT
• only RREQ_RETRIES attempts at TTL=NET_DIAMETER
• Route Interruption• Informs using RERR
• Performs local repair or source initiates a new route discovery
AODV TTL Simulation Parameters Results Main
20/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Effect of the Initial Value of TTL Field
• When destination is closer than the previously known hop count
• The destination was previously two hops away
• The shaded nodes are those which have transmitted a RREQ packet
• Left: the initial value of the TTL field is (2 + TTL_INCREMENT = 4)
• Right: The initial value of the TTL field is TTL_START i.e. 1
S
I4
I5
I6
I3
I1
I2
D
I7
I8
S
I4
I5
I6
I3
I1
I2
D
I7
I8
AODV TTL Simulation Parameters Results Main
21/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Effect of the Initial Value of TTL Field
• When destination is at the same distance as the previously known hop count
• The destination was previously two hops away
• The shaded nodes are those which have transmitted a RREQ packet
• Left: the initial value of the TTL field is (2 + TTL_INCREMENT = 4)
• Right: The initial value of the TTL field is TTL_START i.e. 1
• Requires another phase with expanded ring TTL+=TTL_INCREMENT=3
S
D
I5
I6
I3
I1
I2
I4
I7
I8
S
D
I5
I6
I3
I1
I2
I4
I7
I8
AODV TTL Simulation Parameters Results Main
22/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Simulation Parameters
• OPNET Modelere with wireless suite
• SMP machine with 2 Intel Xeon 3.0 GHz processor
• 2 GB RAM
• Microsoft Windows Server 2003
• Simulation run duration: 1800 seconds
• 1024 Bytes per packet
• Every combination of settings repeated with 5 different seeds
• Random waypoint mobility traces are first evaluated to avoid• Density wave
• Speed decay
AODV TTL Simulation Parameters Results Main
23/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Simulation Parameters
• Node coverage ≈ 250m (radius)• Transmit power = 0.04 watt
• Packet Reception-Power Threshold = 73 dBm
AODV TTL Simulation Parameters Results Main
Simulation Environment
NetworkSize
GeographicalArea
NodeDensity
(per sq. km)
NetworkDiameter(nodes)
NeighborCount
25 nodes800 m
X800 m
39.06 4.52 7.67
100 nodes2000 m
X500 m
100 8.25 19.63
24/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Simulation Parameters
• AODV settingsAODV Parameters
Parameter Value
RREQ_RETRIES 3
ACTIVE_ROUTE_TIMEOUT 3 seconds
DELETE_PERIOD 15 seconds
HELLO_INTERVAL 1 second
ALLOWED_HELLO_LOSS 2
NET_DIAMETER 20
NODE_TRAVERSAL_TIME 0.04 second
TIMEOUT_BUFFER 2
TTL_START 1
TTL_INCREMENT 2
TTL_THRESHOLD 7
LOCAL_ADD_TTL 2
AODV TTL Simulation Parameters Results Main
25/17IBR, TU BraunschweigInconsistent Reactive Routing Components
MAC Overhead
AODV TTL Simulation Parameters Results Main
26/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Routing Overhead
AODV TTL Simulation Parameters Results Main
27/17IBR, TU BraunschweigInconsistent Reactive Routing Components
Data Packet Delivery Ratio
AODV TTL Simulation Parameters Results Main