Multicasting in Ad-Hoc Networks: Comparing AODV and ODMRP
Thomas Kunz and Ed Cheng
Systems and Computer Engineering
Carleton University
http://kunz-pc.sce.carleton.ca/
tkunz@sce.carleton.ca

Mobile Ad Hoc Networks
Infrastructure-less, may need to traverse multiple wireless links to reach a destination

Motivation
Many envisioned applications for ad hoc networks require one-to-many communication
Multicast protocols are intended to efficiently support such communication patterns
Multicasting well researched in fixed networks (i.e., the Internet), but host and router mobility cause problems
MANET specific protocols are being proposed
MAODV: multicast extensions for AODV, establishes shared tree
ODMRP: new multicast protocol, based on per-source mesh
Goal: study and compare protocols to identify possible avenues for improvement

MAODV
MAODV: Multicast Ad Hoc On-Demand Distance Vector protocol
First node to join a group becomes leader
Leader periodically broadcasts group hello messages (including updated group sequence number)
Multicast tree based on hard state, nodes joining or leaving require action to reconfigure the tree
Downstream nodes who detect link failure will try to reconnect to tree
In case of network partition, two trees get established, after network partitions merge, multicast trees are merged again as well.
One node will receive two group hello messages for some multicast group and will ask the leader with lower ID for permission to reconnect and will do so by joining multicast group

ODMRP
ODMRP: On-demand Multicast Routing Protocol
Each source periodically broadcasts join requests, interested receivers reply, mesh gets established

Qualitative Comparison

Simulation Environment
NS2
Widely used network simulator
Simple physical model (free space propagation plus two-way ground reflection)
MAC: 802.11 RTS/CTS
Provides support for node mobility, unicast protocols such as DSR, AODV for ad hoc networks already implemented
Simulation parameters
1000 x 1000 meter area, 250 m radio range, 2 Mbps link capacity
50 nodes, 1 multicast group
900 seconds simulation time
Mobility model: Random Waypoint model, 0 seconds pause time, max speed between 1 m/s to 20 m/s
Traffic: CBR (4 packets of 512 bytes per second and sender)
Parameters varied: Number of Senders, Node Mobility, Group Size

Simulation Environment:
Evaluation Criteria
Packet Delivery Ratio: The ratio of the number of packets actually delivered to the destinations versus the number of data packets supposed to be received
Number of data packets transmitted per data packet delivered: includes retransmissions and dropped packets
Number of control packets transmitted per data packet delivered: routing protocol overhead, normalized by user traffic (protocols are on-demand)
Number of control packets and data packets transmitted per data packet delivered: This measure tries to capture a protocol’s channel access efficiency, as the cost of channel access is high in contention-based link layers

Results (Example): Number of Senders
Twenty group members, maximum speed 1m/s

Results (Example): Number of Senders (cont)

Conclusions
MAODV has poorer packet delivery ratio
MAODV uses a shared tree for data dissemination. If a single tree link breaks because of node movement, packet collision, or congestion, destinations cannot receive packets
ODMRP provides redundant routes with a mesh topology and the chances of packet delivery to destinations remain high even when the primary routes are unavailable
ODMRP suffers from scalability issues as the multicast group increases or the sender size increases
ODMRP maintains per-source meshes connecting receivers and senders. As the number of senders increase, periodic Join Query packets increase, causing higher amounts of congestion and control overhead
MAODV uses a single multicast group leader to send out periodic Hellos through the network.  Increasing the number of senders has minimal impact

Future Work
Study impact of traffic (not just CBR)
Preliminary results show that MAODV is more sensitive to traffic type
Reduce flooding overhead inherent in both protocols using pruning and dominant pruning
MAODV: lower data delivery ratio, but also lower overheads
Improve tree maintenance by pro-actively predicting link failure and triggering tree maintenance BEFORE receivers get disconnected
Results for unicast case show that link breakages based on received signal power strength can accurately be predicted and used to significantly reduce number of dropped packets
Reduce packet drop rates by 30% - 45%
Increase control messages by 19% - 43%
Reduce packet latency by up to 25% (though some increase possible at low mobility rates)