Title: Huayi Wu and Xiaohua Jia
1QoS Multicast Routing by Using Multiple
Paths/Trees in Wireless Ad Hoc Networks
- Huayi Wu and Xiaohua Jia
- Ad Hoc Networks, July 2007
- speaker Yu-Hsun Chen
2Outline
- Introduction
- Related Works
- Problem Formulation
- Multiple Paths/Trees Multicast Routing Protocol
- Three Multiple Paths/Trees Construction
Strategies - Simulations
- Conclusion
3Introduction 1
- Many applications require QoS multicast
- Multimedia group meeting
- Real-time multimedia data dissemination
- QoS multicast routing
- Find a multicast tree
- Rooted from the source node
- Spanning to all destination nodes
- Every path from the source to the destination
satisfies the QoS requirements - Limitation
- Limited bandwidth
4Introduction 2
- The QoS multicast routing in this paper
- Utilize multiple parallel paths or trees to meet
the bandwidth and delay requirement of a QoS call - Major advantages
- Reduce the system blocking probability
- Distributed scheme
- Similar to traditional on-demand routing protocols
5Outline
- Introduction
- Related Works
- Problem Formulation
- Multiple Paths/Trees Multicast Routing Protocol
- Three Multiple Paths/Trees Construction
Strategies - Simulations
- Conclusion
6On-Demand Multicast Routing Protocol 1
- S. J. Lee, W. Su, and M. Gerla, On-Demand
Multicast Routing Protocol in Multihop Wireless
Mobile Networks, Mobile Networks and
Applications, 2002 - ODMRP is a mesh-based, rather than a tree-based
multicast scheme - Use a forwarding group concept
- The mesh provides richer connectivity compared
with trees
7On-Demand Multicast Routing Protocol 2
8Multicast Operation of the Ad-hoc On-Demand
Distance Vector Routing Protocol 1
- E. M. Royer and C. E. Perkins, Multicast
Operation of the Ad-hoc On-Demand Distance Vector
Routing Protocol, IEEE/ACM International
Conference on Mobile Computing and Networking,
1999 - Extend AODV to offer multicast capabilities
- J_flag in RREQ
- Group Hello Messages
- Multicast Activation (MACT)
9Multicast Operation of the Ad-hoc On-Demand
Distance Vector Routing Protocol 2
10A Robust Multicast Routing Protocol
- G. H. Lynn and T. F. Znati, RoMR A Robust
Multicast Routing Protocol for Ad-Hoc Networks,
IEEE LCN, 2001 - Monitor the link and determine the percentage of
time the link was available - Build multiple reliable multicast trees
- m reliable multicast trees
- k edges in common
11Outline
- Introduction
- Related Works
- Problem Formulation
- Multiple Paths/Trees Multicast Routing Protocol
- Three Multiple Paths/Trees Construction
Strategies - Simulations
- Conclusion
12CDMA/TDMA and timeslot assignment
- MAC layer CDMA-over-TDMA channel model
- Conflict free fashion
- The number of timeslots ? free bandwidth over a
link
- 2 free bandwidth of link AB
- Assigning free timeslots to a connection to
- maximize the available bandwidth of the
- connection is NP-hard
13Problem Statement 1
- Definition 1
- The available bandwidth of a link l
- Definition 2
- The bandwidth of a path P
- Definition 3
- The available bandwidth of a tree T
14Problem Statement 2
- Definition 4
- The network cost of a path P
- Definition 5
- The network cost of a multicast tree T
- Definition 6
- The delay of a tree, d(T), is the number of hops
from the root to the farthest leaf node in T
15Problem Statement 3
- Problem
- Given a QoS multicast request
- Set up a multicast connection
- Using multiple paths/trees
- Minimize the network cost
- Meet the bandwidth and delay requirement
16Outline
- Introduction
- Related Works
- Problem Formulation
- Multiple Paths/Trees Multicast Routing Protocol
- Three Multiple Paths/Trees Construction
Strategies - Simulations
- Conclusion
17Route Discovery and Reply Phases 1
- Find candidate paths in parallel as many as
possible - Route discovery
- The source node floods the RREQ packets on-demand
- source, destination-list, seq-ID, type, route,
freeslots, b, TTL - When receiving a RREQ
- Check if there is any common free timeslots
- Append its own address to the route field and
its free timeslot to the freeslots field - Decrease TTL by 1
- Record duplicate RREQ but will not reflood
18Route Discovery and Reply Phases 2
- Route Reply
- The RREP packet
- source, node-ID, type, route, freeslots, b, TTL
- A non-destination node also needs to reply RREP
- Nodes wait either for a pre-specified timeout or
the reception of a certain number of RREQs - Send back a RREP including all the information
about multiple paths going through the node
19Route Discovery and Reply Phases 3
20Maintenance Phases
- Route setup a new node join
- Flood a JO (join) packet
- Route prune a destination node leave
- Send PU (prune) packet to all the upstream nodes
of it - Route recovery
- Destination node is failed
- Its upstream nodes will detect it
- Forwarding node is failed
- Its downstream node floods JO for a new route
21Outline
- Introduction
- Related Works
- Problem Formulation
- Multiple Paths/Trees Multicast Routing Protocol
- Three Multiple Paths/Trees Construction
Strategies - Simulations
- Conclusion
22Shortest Path Tree Based Multiple-paths 1
- To find a shortest delay multicast routes
- The source node can easily find the SPT to all
the destinations - When bandwidth is not enough
- The source selects multiple paths parallel to the
path-segment - When there is a branch node
- Split the path-segment into two parts from the
branch node
23Shortest Path Tree Based Multiple-Paths 2
24Least Cost Tree Based Multiple-Paths 1
- To find a least cost route
- The destination node is added first if it has the
least network cost - When bandwidth is not enough
- Find multiple parallel paths
- The multiple parallel paths are sorted by their
network costs
25Least Cost Tree Based Multiple-Paths 2
26Multiple Least Cost Trees 1
- Find multiple trees directly
- Find a LCT first
- Those links that has no more bandwidth are
removed - The source will keep on searching for a LCT in
the new network graph
27Multiple Least Cost Trees 2
28Outline
- Introduction
- Related Works
- Problem Formulation
- Multiple Paths/Trees Multicast Routing Protocol
- Three Multiple Paths/Trees Construction
Strategies - Simulations
- Conclusion
29Simulation Setup
- Range 100 100
- Number of nodes 100
- Transmission range 30
- Number of timeslots 16
- Network load 0 1
- Bandwidth requirement 2, 4, and 6 timeslots
30Simulation Results 1
- Network cost and success ratio versus multicast
group size
- The MLCT and LCTM perform better in network cost
- Multiple paths/trees greatly increases the
success ratio
31Simulation Results 2
- Success ratio versus network load
- The success ratio of SGT reaches the zero more
quickly with the increase of network load - The success ratio decreases sharply when the
network load reaches a certain threshold
32Simulation Results 3
- Network cost versus network load
- The network cost is higher by using SPTM
33Conclusion
- Advantages of the protocol
- Reduce the system blocking probability
- Distributed scheme
- Similar to traditional on-demand routing
protocols - Comments
- Delay jitter might be an issue
- Network environment