ECE802 Presentation

1
ECE802 Presentation

• Farshad A. Samimi
• Farshad_at_cse.msu.edu
• Spring 2004

Paper A Dynamic Core Based Multicast Routing
Protocol for Ad Hoc Wireless Networks, Subir
Kumar Das, et al., MobiHoc 2002
2
Agenda

• Introduction
• DCMP Protocol
• Mechanism
• Example
• Analysis
• Results (Simulation)
• Conclusions

3
Introduction

• Ad hoc routing protocols
• Table-driven (DSDV, CGSR, WRP)
• Control packets overhead
• Source-initiated on-demand (AODV, DSR, TORA)
• Longer connection setup time
• Multicast vs. multiple unicasts
• Multicast routing protocols for ad hoc networks
• Mesh-based vs. tree-based
• Construction of the multicast tree
• Source-initiated
• Receiver-initiated
• Multicast group maintenance approach
• Soft state
• Hard state

4
Introduction

• Examples of ad hoc multicast routing protocols
• Tree based
• Ad hoc Multicast Routing (AMRoute)
• Assumes existence of a unicast routing protocol
• Creates a virtual bidirectional shared multicast
  tree over the mesh
• Provides robustness in a high mobility
  environment
• Ad hoc Multicast Routing utilizing Increasing
  id-numberS (AMRIS)
• On-demand, source-initiated, and shared tree
  based
• Mesh based
• On Demand Multicast Routing Protocol (ODMRP)
• Uses a forwarding group concept
• Source-initiated and soft state maintenance
  approach
• Flooding of control packets results in
  considerable control overhead
• Core-Assisted Mesh Protocol (CAMP)
• Uses CBT idea to form a mesh with multiple cores
• Eliminated flooding of control packets by using
  core nodes in the mesh
• Depends on the underlying unicast routing
  protocol

5
Introduction

• Motivation
• ODMRP
• Mesh-based, source-initiated
• High packet delivery ratio even at high mobility
• Mechanism
• Robustness at the expense of increased control
  overhead, especially when the number of sources
  is large
• DCMP paper proposal
• Dynamic Core based Multicast routing Protocol
• Builds and maintains a shared mesh a mesh formed
  by a group of core based trees (CBT)
• Improves the scalability
• DCMP vs. CAMP
• CAMP is receiver-initiated
• CAMP depends on the underlying unicast routing
  protocol

6
DCMP Protocol Mechanism

• Sources
• Active, Core Active (Core), Passive, ToBePassive
• Control packets
• JoinReq, Reply, PassReq, Confirm
• Tables
• PassSourceAddr, ConfirmRouteFind
• Parameters
• MaxPassSize, MaxHop, CoreAcceptance flag, Node
  identification number (ID), Forwarding flag
  (FgFlag), ConfirmWait timer, ConfirmRouteDelete
  timer, PassiveSupported counter,
  PassiveSourceExistence timer
• Active to passive status change upon reception of
  a JoinReq
• ToBePassive conditions
• CoreAcceptance flag is set
• Hop distance traveled by JoinReq lt MaxHop
• Node ID of JoinReq Receiver lt Node ID of JoinReq
  source

7
DCMP Protocol Example

• DCMP mesh topology

8
DCMP Protocol Example

• Topology change in DCMP due to movement of node S3

9
DCMP Protocol Example

• Mesh topology in ODMRP

• Mesh topology in DCMP

• DCMP vs. ODMRP

10
DCMP Protocol Analysis

• Analysis of reduction in control overhead
• ODMRP control overhead
• DCMP control overhead
• Reduction in control overhead
• The reduction in control overhead is proportional
  to the number of Passive sources (Sp) in the
  group
• Estimation of number of Passive sources

Number of PassReqs and Confirms
Number of JoinReqs
Number of Replies
11
Simulation

• Simulation environment
• GlomoSim
• IEEE 802.11 in DCF mode (2Mbps, transmission rage
  of 250m)
• 50 mobile nodes moving within a 1000m1000m area
• Default mobility 20m/s
• CBR data flow of 512-byte long packets at 10
  packets/sec
• Active sources flood JoinReq packets at intervals
  of 3 seconds
• Multicast group sizes 5 (small) and 20 (large)
• Number of sources 5
• Duration 200 seconds
• Results averaged over 20 simulation runs
• Metrics
• Number of control packets transmitted per data
  packet delivered
• Represent the degree of control overhead
• Number of data packets transmitted per data
  packets delivered
• Represents the multicast routing efficiency
• Data packet delivery ratio

12
Simulation Results

• Number of Passive sources vs. MaxHop with
  varying MaxPassSize (for large multicast group)
  FIG.8

• Control overhead vs. MaxHop with varying
  MaxPassSize (for large multicast group) FIG.12

• Impact of MaxHop and MaxPassSize Parameters ?

13
Simulation Results

• Packet delivery vs. MaxHop with varying
  MaxPassSize (for large multicast group) FIG.14

• Based on the results, MaxHop and MaxPassSize have
  been chosen as 2 for the rest of the experiments

14
Simulation Results

• Comparison of control overhead for large
  multicast group FIG.16

• Comparison of control overhead for small
  multicast group FIG.15

• Impact of Number of Sources
• Control Overhead

15
Simulation Results
Large number of forwarding nodes keeps the
difference constant

• Comparison of number of data transmissions per
  data packet delivered for small multicast group
  FIG.17

• Comparison of number of data transmissions per
  data packet delivered for large multicast group
  FIG.18

• Impact of Number of Sources
• Multicast Routing Efficiency

16
Simulation Results
DCMP causes fewer collisions but ODMRP provides
more reliability

• Comparison of packet delivery ratio for small
  multicast group FIG.19

• Comparison of packet delivery ratio for large
  multicast group FIG.20

• Impact of Number of Sources
• Data Packet Delivery Ratio

17
Simulation Results
Soft state approach keeps the overhead constant
even in higher mobility

• Comparison of control overhead for small
  multicast group FIG.21

• Comparison of control overhead for large
  multicast group FIG.22

• Impact of Mobility
• Control Overhead

18
Simulation Results
Packet dropping
Frequent link breakings results in a large
number of forwarding nodes

• Comparison number of data transmissions for
  small multicast group FIG.23

• Comparison number of data transmissions for
  large multicast group FIG.24

• Impact of Mobility
• Multicast Routing Efficiency

19
Simulation Results
Larger number of forwarding nodes

• Comparison of packet delivery ratio for small
  multicast group FIG.25

• Comparison of packet delivery ratio for large
  multicast group FIG.26

• Impact of Mobility
• Data Packet Delivery Ratio

20
Simulation Results
Fewer collisions

• Comparison of packet delivery ratio for small
  multicast group FIG.27

• Comparison of packet delivery ratio for large
  multicast group FIG.28

• Impact of Load
• Data Packet Delivery Ratio

21
Summary Conclusions

• Summary
• DCMP is a shared mesh based, on-demand multicast
  protocol for ad hoc networks
• The key concept in DCMP is to make some sources
  Passive, which then forward data packets through
  core active nodes
• DCMP reduces control overhead
• Conclusion
• DCMP increases scalability
• Simulation results
• 30 reduction in control overhead
• 10-15 improvement in multicast routing
  efficiency
• 2 reduction in packet delivery ratio for light
  network loads but improved at high load