Download presentation
Presentation is loading. Please wait.
1
9/25/2000UCLA CSD Gerla, Kwon and Pei On Demand Routing in Large Ad Hoc Wireless Networks With Passive Clustering Mario Gerla, Taek Jin Kwon and Guangyu Pei Computer Science Department University of California, Los Angeles Los Angeles, CA, 90095
2
9/25/2000UCLA CSD Gerla, Kwon and Pei Clustering in Ad hoc Networks A natural way to provide some “structure” in an ad hoc network Better Channel Efficiency(code diversity) Bandwidth allocation & QoS support Cluster based routing -> scalability Suppress redundant transmissions in On- Demand Routing
3
9/25/2000UCLA CSD Gerla, Kwon and Pei Example of Clustering 3 5 6 7 8 1 4
4
9/25/2000UCLA CSD Gerla, Kwon and Pei AODV: flooding O/H AODV requires flood-search to find and establish routes Flood-search: each node forwards Query pkt (RREQ) to neighbors If network is “dense” (ie, several nodes within the tx range), this leads to a lot of redundant transmissions Energy waste & throughput loss
5
9/25/2000UCLA CSD Gerla, Kwon and Pei Clustering helps On-demand routing The network is organized in clusters All nodes in a cluster can communicate directly (one hop) with clusterhead Gateways maintain communications between clusters Only clusterheads and gateways forward search-flood queries Suppress redundant transmissions!
6
9/25/2000UCLA CSD Gerla, Kwon and Pei Example of Clustehead & Gateway Forwarding 3 5 6 7 8 1 4
7
9/25/2000UCLA CSD Gerla, Kwon and Pei Drawbacks of Conventional Clustering (eg,Least ID #) Periodic neighbor connectivity monitoring may lead to high O/H Periodic control traffic not desirable in military covert operations Unstable behavior of “least ID cluster election” scheme: small move -> large change!
8
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering Goals: no monitoring O/H, more stable.. Approach: (a) No “Active” Control Packets: Cluster state information piggybacked on data packets (b) Clusters are built only when on-demand routes are opened (c) Soft state: when data transmissions cease, time-out clears stale clusters
9
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering: example 3 6 7 8 1 4 Assume Node 1 initiates a search flood…. 2 9
10
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
11
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 Clusterhead_ready 9 2
12
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 Clusterhead 9 2
13
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 Ordinary Node 2 9
14
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 9 2
15
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 9 6 7 8 1 4 2
16
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
17
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 9
18
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
19
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 Gateway 2 9
20
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 9 2
21
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
22
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
23
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 9 2
24
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
25
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
26
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 9 2
27
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
28
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
29
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 2 9
30
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering 3 6 7 8 1 4 Resulting cluster structure. 2 9
31
9/25/2000UCLA CSD Gerla, Kwon and Pei Lowest ID Clustering result 3 6 7 8 1 4 3 isolated clouds – 1, 2, and the rest 2 9
32
9/25/2000UCLA CSD Gerla, Kwon and Pei Simulation Environment (GloMoSim) 100 nodes in 1000m x 1000m Transmission range : 150m Mobility model: Random Waypoint AODV unicast routing Random Source/Destination Pairs CBR traffic. 512 bytes per packet, 0.4 packets per sec
33
9/25/2000UCLA CSD Gerla, Kwon and Pei Normalized Routing Overhead
34
9/25/2000UCLA CSD Gerla, Kwon and Pei Mean End-to-End Delay
35
9/25/2000UCLA CSD Gerla, Kwon and Pei Mean End-to-End Delay
36
9/25/2000UCLA CSD Gerla, Kwon and Pei Throughput
37
9/25/2000UCLA CSD Gerla, Kwon and Pei Throughput
38
9/25/2000UCLA CSD Gerla, Kwon and Pei Summary Passive clustering Realistic, “overhead free” mechanism First Declaration Wins rule Stable clusterhead election AODV application Efficient search-flood; higher thoughput; Next: try Passive Clustering on DSR, ODMRP and other search-flood schemes
39
Thank You!
40
9/25/2000UCLA CSD Gerla, Kwon and Pei Chain Reaction (contd) 3 5 6 7 8 1 4
41
9/25/2000UCLA CSD Gerla, Kwon and Pei Chain Reaction (contd) 3 5 6 7 8 1 4
42
9/25/2000UCLA CSD Gerla, Kwon and Pei Chain Reaction (contd) 3 5 6 7 8 1 4
43
9/25/2000UCLA CSD Gerla, Kwon and Pei Chain Reaction (contd) 3 5 6 7 8 1 4
44
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering Pros and Cons Little line overhead ↔ Longer Convergence time Free Neighbor info. ↔ Partial Neighbor Info. Better Structure Easy to Implement Energy Efficiency Continued..
45
9/25/2000UCLA CSD Gerla, Kwon and Pei AODV (Ad Hoc On Demand DV) Routing application AODV version with Hello messages Hello messages exchanged every 1.5 seconds Hello message reduction No Hello if the node is Ordinary node RREQ, RREP, REER cancel scheduled Hello Reduced Flooding Ordinary nodes do not forward the RREQ packets
46
9/25/2000UCLA CSD Gerla, Kwon and Pei Passive Clustering features Passive clustering with 802.11 Data traffic activated process Clusterhead election rule – FDW Cluster time out : 2 sec
47
9/25/2000UCLA CSD Gerla, Kwon and Pei Mean End-to-End Delay
48
9/25/2000UCLA CSD Gerla, Kwon and Pei Chain Reaction set off by motion of node 1 3 5 6 7 8 1 4
49
9/25/2000UCLA CSD Gerla, Kwon and Pei Final Clusters very different from the initial ones 3 5 6 7 8 1 4
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.