Efficient Flooding in Ad hoc Networks using On-demand (Passive) Cluster Formation

1. Efficient Flooding in Ad hoc Networks using On-demand (Passive) Cluster Formation 2003. 04. 18 ICNS Lab Na Gajin

2. Contents • Introduction • Blind Flooding / Efficient Flooding • Related Works and Motivations • Clustering in Ad hoc Networks • Passive Clustering • Simulation Studies • Conclusion

3. Introduction • Mobile ad hoc networks (MANET) • Self-creating, self-organizing and self-administrating without deploying any kind of infrastructure • Wide application in military, commercial and educational environments where fixed infrastructure is not easily acquired • Two nodes communicate directly or via a multi-hop route with the cooperation of other nodes • To find a multi-hop path to another nodes, each MANET node widely use flooding or broadcast

4. Introduction • Route Discovery (DSR protocol) • Route Maintenance “A” “A,B” “A,B,C” A B C D id=2 id=2 id=2 destination source A B C D A B C D A B C D A Hello messages

5. Introduction • Blind Flooding • Node transmits a message to all of its neighbors • The neighbors in turn relay to their neighbors and so on until the message has been propagated to the entire network. • Neighbor degree gets higher, the blind flooding suffers from the increases of • redundant and superfluous packets • the probability of collision • congestion of wireless medium

6. Blind Flooding redundant and superfluous packets

7. Introduction • Efficient Flooding • Only a subset of neighbors is required to participate in flooding to guarantee the complete flooding • In MANET, collecting topological information is very difficult (huge extra overhead) • Many on-demand ad hoc routing schemes and service discovery protocols simply use blind flooding

8. Introduction • New mechanism for efficient flooding suitable for on-demand protocols based on passive clustering • Several contributions with previous flooding mechanisms • It does not need any periodic messages • It does not have any setup latency, and it saves energy with no traffic • Its maintenance is well adaptive to dynamic topology and resource availability changes

9. Related Works • Proposed several heuristics to reduce rebroadcasts • Probabilistic scheme • Rebroadcast the packet with the randomly chosen probability • Counter-based scheme • Rebroadcast if the number of received duplicate packets is less than a threshold • Distance-based scheme • Uses the relative distance between hosts to make the decision • Location-based scheme • Based on pre-acquired location information of neighbors • Cluster-based scheme • Only cluster heads and gateways forward again

10. Related Works • Another approach : exploit topological information • Self-pruning • Each forwarding node piggybacks the list of neighbors of itself on outgoing packet • Dominant-pruning • Extends the range of neighbor information to two-hop away neighbors • Still depend on the periodic hello messages to collect topological information • Extra hello messages consume resources and drop the network throughput in MANETs

11. Motivations • Clustering • Reducing the routing table size • Reducing the communication overhead • Stabling network topology • Be ease of location management • Providing a simple and feasible power control mechanisms

12. Brief Overview of Clustering • Clustering : Another method to select forwarding nodes • Cluster head : representative of each group (cluster) • Gateway : a node belongings to more than two clusters at the same time • Ordinary nodes • Transmission area (radio range) of the cluster head defines a cluster • k-hop clustering

13. Efficient Flooding with Clustering CLUSTER HEAD GATEWAY S ORDINARY NODE Flooding Only cluster heads and gateways rebroadcast and ordinary nodes stop forwarding

14. Motivations • Clustering in ad hoc networks • Hierarchical routing schemes • Master election algorithms • Power control • Reliable and efficient broadcast • Cluster architecture commonly not used • Previous clustering schemes are based on the complete knowledge of neighbors • None of the clustering algorithms has proposed a gateway reduction mechanism to select the minimal number of gateways • The previous clustering requires huge maintenance cost in high mobility

15. Lowest-ID Clustering • Each node is assigned a distinct ID • Periodically, the node broadcasts the list of nodes that it can hear • A node which only hears nodes with ID higher than itself is a “clusterhead” Clusterhead Gateway Ordinary Node

16. MPR (Multipoint Relays) • Reduce the flooding of broadcast messages • Set of one-hop neighbors and two-hop neighbors • To get the information about the one-hop neighbors, most protocols use some form of HELLO messages periodically

17. Three Important Observations • The selection mechanism to choose the dominant set should be efficient and dynamic • In a MANET, collecting accurate topological information is very hard and carries the huge overhead • Clustering schemes is independent of the network topology  With keeping advantages of clustering, our scheme eliminates the main control overhead

18. Overview of Passive Clustering • On-demand protocol • Constructs and maintains the cluster architecture only when there are on-going data packets that piggyback “cluster-related information” • Each node collets neighbor information through promiscuous packet receptions • First Declaration Wins • Node that first claims to be a cluster head “rules” the rest of nodes in its clustered area • Gateway Selection Heuristic • Elect the minimal number of gateways

19. Construction and Maintenance • The IP option field for cluster information • Node ID : the IP address of the sender node • State of cluster : the cluster state of the sender node • If a sender node is a gateway, then it tags two IP addresses of cluster heads which are reachable from the gateway • Initially joining node or floating node sets the cluster state to INITIAL

20. Passive Clustering Algorithm • Cluster states • INITIAL, CLUSTER_HEAD, ORDINARY_NODE, GATEWAY, CH_READY, GW_READY and DIST_GW • Packet handling • Upon sending a packet, each node piggybacks cluster-related information • Upon a promiscuous packet reception, each node extracts cluster-related information of neighbors and updates neighbor information table

21. Passive Clustering Algorithm • A cluster head declaration • INITIAL  CH_READY : packet arrives from another node that is not a cluster head • With outgoing packet, a CH_READY node can declare as a cluster head • Becoming a member • A node becomes a member of a cluster once it has heard or overheard a message from any cluster head. • A member node can serve as a gateway or an ordinary node depending on the collected neighbor information • A member node can settle as an ordinary node only after it has learned enough neighbor gateways

22. Gateway Selection Heuristic • Gateway : a bridge node that connects two adjacent clusters • Only one gateway is needed for the each pair of two adjacent clusters • Gateway selection mechanism that eventually allows only one gateway for each pair of two neighboring cluster heads

23. Gateway Selection Heuristic • candidate gateway : a node belongs to more than two clusters at the same time • If the node finds 2 cluster heads, then it finalizes its role as a gateway and announces 2 cluster heads to neighbors • If a gateway has received a packet from another gateway which has announced the same pair of CHs, then this node compares the node ID of itself with that of the sender. • If this node has the lower ID, it keeps its role as the gateway. • Otherwise it changes the pair of CHs or changes its state

24. Gateway Selection Heuristic 4 1 CLUSTER HEAD (1,4) GATEWAY ORDINARY NODE (1,7) (4,5) GW_READY NODE 5 (5,7) 7 There is at most one gateway between any pair of two cluster heads

25. Simulation Studies • Flooding efficiency with passive clustering • Apply passive clustering to representative reactive ad hoc routing protocols (AODV, DSR) • Flooding Experiments • TNP (the Total Number of Packets sent for one broadcast) • NDB (the Number of nodes Delivered the Broadcast) • BF (Blind Flooding) • MPR-F (Flooding with MPR scheme) • AC_LID-F (Flooding with active clustering with Lowest Id ) • PC_LID-F (Flooding with passive clustering)

26. Flooding Experiments • Fixed network size with node mobility (100 nodes) Total Number of Packet sent Delivery Count (NDB)

27. On-demand Routing -AODV Normalized Control Overhead Delivery Ratio

28. On-demand Routing - DSR Normalized Control Overhead Delivery Ratio

29. Conclusion • Passive Clustering protocol • Effective gateway selection heuristic • Efficient flooding based on topological information • Applicability of passive clustering to a few reactive routing protocols

30. References • Taek Jin Kwon  ,  Mario Gerla, “Efficient Flooding in Ad hoc Networks using On-Demand (Passive) Cluster Formation”, ACM SIGCOMM Computer Communication Review ,2002. • Gerla, M.; Taek Jin Kwon; Pei, G., ”On-demand routing in large ad hoc wireless networks with passive clustering”, Wireless Communications and Networking Conference, 2000. • Amir Qayyum, Laurent Viennot, Anis Laouiti, “Multipoint relaying: An ecient technique for flooding in mobile wireless networks”, Technical Report RR-3898, INRIA, February 2000.