Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

17
1 Efficient flooding with Passive clustering (PC) in Ad Hoc Networks Taek Jin Kwon, Mario Geria University of California at Los Angeles ACM SIGCOMM 2002

description

Efficient flooding with Passive clustering (PC) in Ad Hoc Networks. Taek Jin Kwon, Mario Geria University of California at Los Angeles ACM SIGCOMM 2002. outline. Introduction Limitations of existing clustering schemes Passive clustering Simulation Conclusion. introduction. CONCEPT - PowerPoint PPT Presentation

Transcript of Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

Page 1: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

1

Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

Taek Jin Kwon, Mario GeriaUniversity of California at Los AngelesACM SIGCOMM 2002

Page 2: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

2

outline

Introduction Limitations of existing clustering

schemes Passive clustering Simulation Conclusion

Page 3: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

3

introduction

CONCEPT Flood search is the capstone of all on-

demand routing protocols. Although some mechanism like

“duplicates are detected” can reduce some overhead.

Flooding generates replicated packet arrivals to each node.

Cluster is a good idea to reduce flooding overhead.

Page 4: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

4

Limitations of existing clustering schemes

Node locations and neighborhood information are key for clustering; unfortunately, they do vary in time.

In the period of neighbor learning and initial clustering, it is essential that there is no mobility for proper convergence.

The cluster structure is useful only after the cluster have been created.

Page 5: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

5

Passive clustering

On-demand cluster. Create by route request. Utilize the advantages of cluster on the

fly.

Create a sub-layer under MAC layer for passing the packet for this node.

Page 6: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

6

Passive clustering

Add 2-bits filed in MAC header for 4 states. INITIAL CLUSTERHEAD ORDINARY_NODE GATEWAY embedding/stripping the 4 states in CSL. And another internal state

“CLUSERHEAD-READY”

MAC

Cluster sub layer (CSL)

IP

Page 7: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

7

Passive clustering Timing of states

Initial: At cold start The number of this non-clusterhead node’s neighbor head is zero 。

Clusterhead: At cold start, if the first received MAC packet is not send by header then set itse

lf “CLUSERHEAD-READY” and then set itself “CLUSTERHEAD” when forwarding.

ORDINARY_NODE: Anytime when received a message which send by header.

Gateway: When a non-clusterhead node hears a packet from a clusterhead, the node be

comes a gateway if α*NC + β > NG. NC = the number of cluster-head. NG = the number of neighboring gateway. β, α = user define parameter

Page 8: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

8

Passive clustering An example to show “on the fly”:

Source send the RREQ. When the neighbor of source receive the RRE

Q, change it’s state to ”Clusterhead-Ready.”

When any of above node want to forward RREQ, set it’s state to ”CLUSTERHEAD” and broadcast it.

The header selection schema is “first declaration wins.”

Page 9: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

9

Passive clustering At this time, all nodes received the clusterhead announcement ca

n changed to gateway (α*NC + β > NG). α set to 1 and β set to 0. Now NC=1 and NG =0.

SD

RREQ

RREQ+head stateRREQ+gateway stateRREQ+gateway state

silence

silence

RREQ+head state

Find D

Clusterhead-Ready Gateway

Clusterhead-Ready

CLUSTERHEADCLUSTERHEAD

Page 10: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

10

Simulation Simulation environment:

Implement in GloMoSim. 100 nodes in 600x600 meters square space. 150m of transmission range. Traffic source are CBR.

One packet per sec. Date packet: 512bytes. Control packet: 32 bytes.

Each simulation duration is ten min. Use Random way point model.

Pause time = 10 sec.

Page 11: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

11

Simulation Test the path-finding (1):

Environment: 2400 random source and destination pairs. Ran from cold start one-by-one.

Only one data packet is sent from each source to each destination.

There is only one source and destination in a given period (which is much larger than cluster time out (=2 seconds)) to ensure that no residual clustering structure remains after the single transmission.

Result: 100% packet delivery was observed with the

experiment.

Page 12: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

12

Simulation Test the path-finding (2):

Environment: Varying nodes speeds of 0,2,4,6,8 meters per

sec. Send a packet every 15 sec for 100 min. (1/15

packet/sec) Slow rate to ensure that the previous cluster

structure dissolves after timeout(2 sec). Result:

The delivery of speeds of 0,2,4 meters/sec is 100%.

Speed of 6 meters/sec is 99.25% and 8 meters/sec is 98.25%.

Page 13: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

13

Simulation

Throughput(kbits/sec)

Throughput(kbits/sec)

Page 14: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

14

Avg

hop

dis

tanc

e

Offered load(Kbps)

Avg

hop

dis

tanc

e

Page 15: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

15

Simulation

Offered load(Kbps)

Num

ber

of f

lood

ing

rela

y

Page 16: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

16

Conclusion Passive cluster can reduce more flooding

overhead than the active flooding schemas.

Passive cluster can solve the limitations of active cluster schema.

Obviously, total flooding will work best in sparse and very dynamic networks

Page 17: Efficient flooding with Passive clustering (PC) in Ad Hoc Networks

17

END Reference:

[1] “On Demand Routing in large AD Hoc wireless Networks with Passive Clustering” (WCNC 2002)

[2] ”Energy Balanced Broadcasting Through Delayed Intelligence” (ITCC 2005)

[3] ”Scalable AODV with Efficient Flooding based on On-Demand Cluster”( Mobile Computing and Communication Review, Volume 6, Number 3)