Load Balancing of Multipath Source Routing in Ad Hoc Networks

Lianfang Zhang, Zenghua Zhao, Yantai Shu, and Lei WangDepartment of Computer ScienceTianjin University, Tianjin 300072, China

OutlineI. Introduction

II. An Overview To MSR

III. Model Based Analysis

IV. Simulation Result

V. Conclusion

I.Introduction(1/2) Mobile ad hoc networks are characterized

by multi-hop wireless links, the absence of any cellular infrastructure, and frequent host mobility

Manage frequent topology changes and need to be bandwidth- and power-efficient

I.Introduction(2/2) DSR

Single path routing under-utilize resources and can not cope with

congestion and link breakage MSR

Probing mechanism Refresh the information in cache, to delete sta

le path and to fined new one in time

II.An Overview to MSR Maintaining alternative paths requires more

routing table space and computational overhead

Source routing is flexible The on-demand nature of DSR can reduce

greatly the routing storage and routing computation

II.A. Path Finding Path information →Route cache To achieve high path independence,

disjoint paths are preferred in MSR. No looping problem

II.B. Probing and Load Balancing Use probing as a feed back control

mechanism Send probing packets periodically to each

path and measure their round-trip time Estimate path delay Distribute traffic over different paths in order

to achieve a minimum mean delay for the whole network

III. Model Based Analysis Intermediate nodes would do nothing

except to forward the packet Adding no more processing complexity

than DSR All path calculation is done in the source

hosts and optimal routing is intimately related to load balancing

III.A. General Analysis on Load Balancing(1/3)

N parallel M/M/1 queues

III.A. General Analysis on Load Balancing(2/3)Objective

Subject to

III.A. General Analysis on Load Balancing(3/3)

(1)

(2)

(3)

III.B. Delay Performance Evaluation in Two-path Case(1/4)

III.B. Delay Performance Evaluation in Two-path Case(2/4) Assuming and

III.B. Delay Performance Evaluation in Two-path Case(3/4)

III.B. Delay Performance Evaluation in Two-path Case(4/4)

Since , after substituting

III.B. Delay Performance Evaluation in Two-path Case

III.C. Discussion on the Heuristic Equation(1/3)

1 21 2 1 2

1 1 1 1x x x x

T T T T

1 1, , 1,2, , ,i j

i j

x x i j N i jT T

Eqn. (6) shows that the difference between the traffic distributed on any two paths must be proportional to the difference of the paths’ average packet.

(6)

III.C. Discussion on the Heuristic Equation (2/3)

max

max

1

, 1,2, ,1k

kk

ddW k Nd

d

refers to the weight of path measured in number of packets to be sent consecutively on the same path every time

kkW

III.C. Discussion on the Heuristic Equation (3/3)

max

1 1 1 1, . .ij i j ij

i j i j

W W W d i e Wd d d d

If we distribute traffic according to the weights, we may achieve near-optimal routing

maxmin , , 1, 2, ,kk

dW U R k N

d

is a factor to control the frequency of switching between routesR is a bound to insure that should not to be too largeU kW

IV. Simulation Result(1/3) CBR

UDP FTP

TCP

IV.B. Simulation Result (2/3)

IV.B. Simulation Result (3/3)

V. Conclusion End-to-end delay Network resource

Future Work Study of optimal load balancing scheme QoS support in MSR