Energy Efficiency of Load
Balancing in MANET Routing
Protocols
Sunsook Jung
Nisar Hundeware
Dr. Alex Zelikovsky
Department of Computer Science
Georgia State University
Outline
 AODV-Node Caching
 AODV-NC with load balancing
 AODV-NC with adaptive work load balancing
 Energy and Routing Efficiency Metrics
 Implementation
 Conclusion
Mobile Ad hoc Networks
 An ad hoc wireless network is an autonomous system
consisting of hosts that do not rely on the presence of any fixed
network infrastructure.
 Applications

battlefield, data acquisition in hostile terrain etc.
 Characteristics of MANET




Infrastructure less mobile networks
All nodes can move and can be connected
dynamically
No fixed routers, no base stations
All nodes are treated as routers
Routing in MANET
 Routing protocols – DSR, DSDV, AODV, TORA
 Objectives of MANET routing
-
to maximize network throughput
to minimize energy consumption
to minimize delay
 Performance Metrics
-
Packet Delivery Ratio
Routing Overhead
End-to-end delay
AODV with Node Caching
 Drawbacks of AODV


Route Request is done using flooding
Result in redundant packet overhead
 Cached node
- have forwarded data packet recently
- have more reliable information about its neighbors and have
better locations
- used to forward RREQ

This is not a broadcast since nodes which are not cached
drop RREQ
-> reduce the Rout Request overhead
Implementation of Node Caching
 Time Threshold H



The Route Request packet has time threshold H
A node N maintain the time T(N) when N forwarded the last
data packet
Upon receiving the Route Request packet by Node N at
current time T
 N check the condition T – H <= T(N)


True : forward the Route Request packet
False: drop the Route Request packet
Route Discovery in AODV-NC
Since T- H > T(N),
RREQ is dropped
S
3
4
9
2
6
RREQ is
forwarded
1
7
5
8
D
Recently sent a data packet
Performance Improvement of AODV-NC
 Relative routing overhead
 Reduced by average 89%
 Delivery ratio
 Increased by average 20%
 End-to-end delay
 Decreased by average 63%
Forwarding Load Balancing

Some specific nodes are overused
in AODV-NC.

To prevent unfairness of node
caching, load balancing scheme
was imposed.

AODV-NC (H : n - t)


n is the maximum number of data
packets forwarded by node N
during time period t
If node N forward cacheconstrained RREQ more than n,
the node N forward only standard
RREQ and data packets during
the break t.
Workload-based Adaptive
Load Balancing
 Forwarding load balancing
algorithm is not self
adaptive.
 Lee et al suggested
workload-based adaptive
load balancing algorithm.
 Drop RREQ according to the
length of the message queue
and the outstanding
workload in nodes.
Energy Efficiency Metrics
 The total energy consumption
 The throughput and network lifetime with limited
energy amount
 The energy usage per delivered packet
 The energy usage per hop
Routing Efficiency Metrics
 Relative routing overhead
 Delivery ratio
 End-to-end delay
 Average number of hops and optimal hops
Optimal hops are calculated by NS2.
 Normalized hops
 average hops/optimal hops
 Distribution of average number of hops

Simulation Study
 All simulations have been performed on NS2 (version ns-
2.26)
 All the scenario files have been generated using NS2

Using setdest and cbrgen.tcl program
 All energy efficiency measured by Energy Model in NS2



TX power: 0.6W, RX power: 0.3W, Idle: 0.1W
Initial energy for energy consumption: 1000J
Initial energy for network lifetime: 300J
Simulation Study (2)
• Parameters for simulations
Grid Area
1000m x 300m
Number of Nodes
50
Communication Range
250m
Type of Connection
CBR / TCP
Packet Size
512 bytes
Pause Time
0 second
Simulation Time
1800 seconds
Packet Rate
4 packet per second
Speed
1, 5, 10, 20 m/s
Connections
10, 20, 30, 40, 50, 60 connections
Results : Energy Consumption
 AODV-NC(1:300-120) uses the least energy in connection 20.
 However, AODV uses the least energy in connection 60.
-> explained by network throughput
Results : Throughput and Network lifetime
 AODV-NC(0.1)-WLB shows the highest throughput both
connection 20 and 60.
 AODV-NC(1:300-120) shows the longest network lifetime.
Results : Energy consumption per packets and hop
 AODV-NC(1:300-120) uses the least energy to deliver a data
packet.
 AODV-NC(1:300-120) and AODV-NC(0.1)-WLB uses the least
energy for one hop.
Results: Routing Efficiency
 Average number of hops

AODV-NC(1:300-120) delivers packets with the
smallest number of hops
 Normalized hops

AODV-NC(0.1) shows the lowest ratio
 Distribution of average hops

AODV-NC sends packets with smaller hops than
AODV.
Results: Routing Efficiency (2)
 For various speeds of nodes




From 1m/s to 20m/s
AODV-NCs with load balancing shows the better
performance than AODV-WLB.
In high mobility, AODV-NC-WLB and AODV-WLB
failed to find a path at the first attempt.
It causes the increments of overhead and delay.
Results: Routing Efficiency (3)
 For various connections of nodes




From 10 to 60 connections
At high workload conditions, WLB improves
performances.
With AODV, WLB improves delivery ratio, relative
overhead and delay up to 6%, 23% and 7.5%
With AODV-NC, WLB improves delivery ratio, relative
overhead and delay up to 32%, 85% and 41%
Conclusion
 With non-adaptive and adaptive load balancing
techniques combines with AODV-NC shows better
performance than AODV itself in energy efficiency as
well as routing efficiency.
 AODV-NC-WLB is the best in network throughput
 AODV-NC-(H:t-n) shows the longest network lifetime
Questions?