You are on page 1of 39

Design and Evaluation of an IEEE 802.

11 Based
Dual MAC for MANETs

Dissertation

submitted in partial ful llment of the requirements


for the degree of
Master of Te hnology

by
Satyajit Rai
Roll No: 01329009

under the guidan e of


Prof. Sridhar Iyer

and

Dr. Leena Chandran Wadia

Kanwal Rekhi S hool of Information Te hnology


Indian Institute of Te hnology, Bombay
Mumbai - 400076.
2003

Dedi ated to
my mother
Smt. Manorama Rai

and
my father
Shri Harish handra Rai

Dissertation Approval Sheet


This is to ertify that the dissertation titled
Design and Evaluation of an IEEE 802.11 Based Dual MAC for MANETs

By
Satyajit Rai

(01329009)

is approved for the degree of Master

of Te hnology.

Prof. Sridhar Iyer


(Guide)

Dr. Leena Chandran Wadia


(Co-Guide)

Internal Examiner

External Examiner

Chairperson
Date :

Abstra t
Multihop ad-ho wireless networks o er great hallenges for proto ol designers. Stations
in su h networks are onstrained by fa tors like low power, limited bandwidth, link
errors, and ollisions. Changes are needed at various levels of the proto ol sta k, most
importantly at the medium a ess layer (MAC). The medium a ess me hanism in
multihop wireless networks should minimize ollisions, and take are of the hidden and
exposed node problems. The IEEE 802.11 MAC with Distributed Coordination Fun tion
(DCF) does not s ale well in su h networks. We introdu e Point Coordination Fun tion
(PCF) in the region of high tra areas, and dis uss its e e t on network performan e.
To improve network s alability and throughput, we propose the design of a new MAC
alled Dual MAC. This work dis usses ar hite ture and working of the dual MAC in
detail. Performan e results of the network using dual MAC are presented, and ompared
with that of pure DCF operation.

iii

A knowledgments
I express my sin ere gratitude towards my guides Prof. Sridhar Iyer and Dr. Leena
Chandran Wadia for their onstant help, en ouragement and inspiration throughout
the proje t work. Without their invaluable guidan e, this work would never have been
a su essful one. I would also like to thank the members of the Mobile Computing
Resear h Group at KReSIT, namely Srinath Perur, Vijay Rajsinghania, Vikram Jamwal,
Deepanshu Shukla, Anupam Goyal, and Abhishek Goliya for their valuable suggestions
and helpful dis ussions. Last, but not the least, I would like to thank the whole KReSIT
family whi h made my stay at IIT Bombay a memorable one.

Satyajit Rai

IIT Bombay
January 15, 2003

iv

Contents

Abstra t

iii

A knowledgments

iv

List of Figures
1

vii

Introdu tion: Cooperative Ad-Ho Networks

1.1
1.2
1.3
1.4

Introdu tion to Wireless Ad Ho Networks . . . . . . . . . .


Multi-hop Wireless Networks . . . . . . . . . . . . . . . . . .
Problems with Centralized Tra . . . . . . . . . . . . . . . .
Requirements and Challenges of Multi-hop Wireless Networks
1.4.1 IP and Routing . . . . . . . . . . . . . . . . . . . . . .
1.4.2 MAC . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.3 Physi al: Hidden and Exposed Node Problems . . . .
1.5 Thesis Obje tive and S ope . . . . . . . . . . . . . . . . . . .
1.6 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . .
MAC in Multi-hop Ad-Ho Networks

2.1 Channel A ess Me hanisms


2.1.1 TDMA . . . . . . .
2.1.2 CSMA/CA . . . . .
2.1.3 Polling MAC . . . .
2.2 Available MACs . . . . . .
2.2.1 IEEE 802.11 MAC .
2.2.2 HiperLan-I . . . . .
2.2.3 HiperLan-II . . . . .
2.3 IEEE 802.11 Operation . .
2.3.1 DCF Operation . . .
2.3.2 PCF Operation . . .
2.3.3 Summary . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

1
1
2
2
2
3
3
4
4

5
5
6
6
7
7
7
8
8
9
10
11

IEEE 802.11 MAC in Multi-hop S enario

12

Dual MAC

15

3.1 Multihop S enario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .


3.2 IEEE 802.11 Operation in Multihop Networks . . . . . . . . . . . . . . .
3.3 Problem Des ription . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1 Need for Dual MAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Overview of Dual MAC . . . . . . . . . . . . . . . . . . . . . . . . . . .

12
13
14

15
15

4.3 Ar hite ture of Dual MAC . . . . . . . . . . . . . . . . . . . . . . . . . .


4.4 Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5 Design Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . .

16
16
17

Dual MAC Implementation in NS

18

5.1 Existing Node Ar hite ture in NS . . . . . . . . . . . . . . . . . . . . . .


5.2 Node Ar hite ture of Dual MAC . . . . . . . . . . . . . . . . . . . . . .
5.3 Implementation in NS . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Simulation and Results

6.1
6.2
6.3
6.4

Simulation Setup . . . . . .
Simple S enario and Results
Generi S enario and Result
Dis ussion on Results . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

18
19
20

22

22
22
24
27

Con lusion and Future Work

28

Bibliography

30

7.1 Con lusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .


7.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vi

28
28
29

List of Figures
1.1 Multihop ooperative Network . . . . . . . . . . . . . . . . . . . . . . .
1.2 Hidden and Exposed Node S enario . . . . . . . . . . . . . . . . . . . .
2.1
2.2
2.3
2.4
2.5
2.6
2.7

TDMA hannel A ess Me hanism


CSMA hannel A ess Me hanism
Polling Me hanism . . . . . . . . .
IEEE 802.11 Ar hite ture . . . . .
DCF a ess using RTS/CTS . . . .
PCF a ess . . . . . . . . . . . . .
PCF Transmissions . . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

2
3

.
.
.
.
.
.
.

5
6
7
9
9
10
11

3.1 Multihop S enario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .


3.2 Hybrid PCF-DCF Operation . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Comparison of DCF and PCF in Multihop Networks . . . . . . . . . . .

12
13
14

4.1 Ar hite ture of Dual Ma . . . . . . . . . . . . . . . . . . . . . . . . . .


4.2 Operation of Dual MAC . . . . . . . . . . . . . . . . . . . . . . . . . . .

16
17

5.1 Ar hite ture of NS Node . . . . . . . . . . . . . . . . . . . . . . . . . . .


5.2 Ar hite ture of Dual Node . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 Dual MAC implementation in NS . . . . . . . . . . . . . . . . . . . . . .

18
19
20

6.1
6.2
6.3
6.4
6.5
6.6
6.7

23
23
24
25
25
26
26

Simple S enario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Throughput omparision of Dual MAC Vs DCF MAC in simple s enario
Pa ket Delivery Ratio for Dual MAC Vs DCF MAC in simple s enario .
Dual MAC Vs DCF MAC at 10 pa kets/se . . . . . . . . . . . . . . . .
Dual MAC Vs DCF MAC at 20 pa kets/se . . . . . . . . . . . . . . . .
Dual MAC Vs DCF MAC at 30 pa kets/se . . . . . . . . . . . . . . . .
Dual MAC Vs DCF MAC - overall Throughput . . . . . . . . . . . . . .

Chapter 1
Introdu tion: Cooperative
Ad-Ho Networks

1.1 Introdu tion to Wireless Ad Ho Networks


In re ent times, the wireless networks have be ome very popular. Wireless LANs are
being deployed on airports, onferen es, et . People have started using portable laptops
to a ess Internet and other resour es using wireless networks while moving. Another
area whi h has generated a lot of interest re ently, is wireless ad-ho networks.
An ad-ho network is formed when two or more stations ome together to form an
independent network. Ad-ho networks are also termed as infrastru ture-less networks
sin e as they do not require any prior infrastru ture. Two stations that are within
transmission range of ea h other are alled one hop neighbours. Multihop ad-ho networks are ones in whi h the stations an talk to stations more than one hop away via
intermediate stations.
Cooperative ad-ho networks are formed by several homogeneous wireless stations.
All the stations ooperate with ea h other, i.e., the tra for the stations that are more
than one hop away is routed by the intermediate stations. The intermediate stations
are alled relaying stations.

1.2 Multi-hop Wireless Networks


Cooperative multihop ad-ho wireless networks onsist of a group of stations onne ted
to ea h other over one or more hops. If two ommuni ating stations are more than one
hop away, the intermediate stations route the pa kets from sour e to destination. Disaster management operations and battalion of soldiers are the example of appli ations
of su h ooperative ad-ho wireless networks. Most often, the tra in these networks
is dire ted towards a entral ontrolling station. The ontrolling entral station is more
apable than other stations and in most ases, an also ommuni ate with headquarter/head o e. The Figure 1.1 shows a multihop ooperative network with the entral
ontrolling station labeled A. The stations B, C, D and F are dire tly onne ted to
station A over a single hop; the stations E, G, H, I and K are onne ted to the station
A over two hops; and the station J is onne ted to station A over three hops.

1.3 Problems with Centralized Tra


..................
....
.....
...
..
..
..
......................
....
....................
..
....
....
.
...
.....
.
.
.
.
.
.
.
..
.
.
.
.
.
...
............................................... ...
...
..
...
.
.
.
.
.
...
.
....
..
.......
.
.
........ ...................
.
....
.........
. ....
.
.
.
....
......
....... ..........
.. ............................
...
.
......
.
.
.
.
.
.
....................
..
.
.
..
.
.
.
.
......
....
....
...
...
...... .......
.
.
.
...
.
.
.
.
.
.....
.
.
.
.
.
.
.
.
.
.
.
.
.
.....
.............
................
.
..
......
.......
.
.
.
.
.
.
.
.
.
..
.
...... ....................
...
.
.
.......
.
.
.
.
.
...
.
.
.
.
.
.
.
.
.
.
....
...........
..
..
....
.
.
.
.
.
.
.
.
.
.
.
.................
...
..
..
.
.
.
..
.
...
..
.
..
...
..
...
...
...
....
...
..
......
..
........
..
...
.. ............... ..........
..
.....
....
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...
.
.
.....
.
.....
.
...
....
.
.
.....
.
.
..
.
.
.
.
.
.
.
.....
...
...
..
..
....
.
.
.
.
.
.
.
.
.
.
.
.
..
..
..... .................
....
.........................
...
.
........
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
....
....
...
.
...
..
...
..
... ................................
.
.
.
.
..
.
...
.
.
..
..................
..
.
.....
.
...
.
.
.
.
.......................
.
.
.
.
.
................
.
...
...
.
..
...
....
.
.
.
.
.
.
.
..
..
...
.
.
.
.
.
....
.
.
.
.
.
..
.
.
.
.
.
.
...
.....
.
......
........
.
..
.................. .......
...
.... ................. ....
..
....
....
.
.
.
.
....
.
.
...
.
....
..
...
....
...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
....
...........
...
....
....
.
.
.
.
....
..
....
....
...
.... .............. ........
..
....
.........
..
....
....
...
...
...
...
..
.
.
.
....................
...
..
.
.....
...
.
.
..
..
...
..
...
.
.
.
..
.
.....
..
.
.
.
...................
..
...
...
..
....
...
.
.
......
.
...............

Figure 1.1: Multihop ooperative Network

1.3 Problems with Centralized Tra


The tra around the entral stations (station A in Figure 1.1) is quite high owing to
reason that most of the stations ommuni ate with this station. In traditional sense,
this model is like a lient-server model where the entral station a ts as a server and
all other stations are lients. The high tra in the entral region auses large number
of ollisions if the hannel a ess me hanism is not designed arefully. This auses
enormous throughput drop in the network. The problems asso iated with the entralized
s enario and their solutions are dis ussed in detail in later hapters.

1.4 Requirements and Challenges of Multi-hop Wireless


Networks
Multihop wireless networks are fa ed by hallenges not present in the wired networks.
Mobility of stations in wireless networks gives rise to issues like route hanges and link
failures. Unlike the wired networks, the main ause of errors in the wireless networks
is due to errors on wireless hannels. These reasons require hanges at various layers of
proto ol sta k.
1.4.1

IP and Routing

The stations in wireless network do not remain in the same subnet due to mobility;
hen e either their IP address needs to be hanged and/or the pa kets be forwarded to
them. These requirements have lead to development of mobile IP[1 where the addresses
are assigned to mobile hosts dynami ally and the pa kets are appropriately forwarded
to them. Frequent link hanges in ad-ho networks ause hange of routes between
stations. This requires that the routing proto ol take are of these hanges and update
the routes frequently. Spe ial routing proto ols like AODV[2, DSR[3 and DSDV[4
have been proposed for use in wireless ad-ho networks. We will not dis uss more about
the IP and routing in this thesis.

1.4 Requirements and Challenges of Multi-hop Wireless Networks


1.4.2

MAC

The link hara teristi s in wireless environments is very di erent from that of wired
networks. At link layer we have to fa e following hallenges:
Bandwidth is the one of the most s ar e resour e in wireless networks.
The available bandwidth in wireless networks (2-10Mbps) is far less than the wired
links (typi ally 100Mbps).

Bandwidth:

The transmission range of stations depends upon the transmitted power


and various sensitivity values. Unlike wired networks all stations on a LAN an
not listen to one another.

Range Issues:

The wireless stations are battery operated and therefore higher transmission
power leads to faster degeneration of the batteries. On the other hand, if we keep
transmission power too small, the stations may no longer be in range of ea h other.

Power:

Sin e all stations an not listen to ea h other, transmission from two station
may lead to ollision at another station.

Collisions:

Channel fading and interferen e ause link errors and these errors may
sometimes be very severe.

Link Errors:

1.4.3

Physi al: Hidden and Exposed Node Problems

The transmission range of stations in wireless network is limited by the transmission


power, therefore, all the station in a LAN an not listen to ea h other. This means that
normal arrier sense me hanism whi h assumes that all stations an listen to ea h other,
fails. In parti ular, this gives rise to hidden node and exposed node problem. Consider
stations A, B, C and D as shown in Figure 1.2. With referen e to the Figure 1.2, hidden
and exposed node problem an be des ribed as follows:
..
.
...
.
...
.
.
....
...
..
...
.
.
.
.
.
.
...
..
....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...
.
.....
..
...
.....
.
.
.
.
.
.
.
.
.
....
..
...
.
...
.
.
.
.
..
..
..
.
..
.
.
.
.
.
..
.
...
.
..
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
..
............
..
............................
.
.
.
.
.
.
.
.
.
..
....
..
..
.
.
....
.
.
....
.
.
.
.
.
.
.
.
.
...
.
..
..
.
..
.
...
.
.
..
.
.
.
.
.
.
....
...
..
.
.
...
.....
.
.
.
...
.
.
.
.
..
.
.
.
.
.........................
.
...
.
..
.............. ....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...
.
.....
...............
.
..
.....
.
.
.
..............
...
.
.
.
..
.
.
.
.
.
.
.
.... ..
..............
...
.
....
...
.............. ..
.
.
.
.
.
.
...
.
.. .
...............
.
..
.
.
....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..............
.....
... ...
.
..
.
..............
.............................
.
.
.
.
.
.
.
.
...............
..........
..
.
. .
..
..
...............
........ ............
..
... ..
.............. ... ..
.....
....
..
..
..
..
................... ..
...
..
....
..
..
..
...
...
..
.........
..
..
....
..
...
...
.
..
.....
...
.
..
....
...
...
.
..
.
.
.
.
.
.
........... ..........
.
.
.
.
.
.....
.
.
.. ......................
.
.
.
..
.
.
......
.
..
..
...........
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
........
..
..
.
. ....................
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
....
..
.. ..........
..............
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
........
.....
.
..
....
...........
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..........
..
......
. .........
..
.............
......
.
..
....
... .... .........................
......
..
.
......
..
......................
....
.
......
..
...
...
......
....
.
.
.
.
.
.
.
.
......
.
.
..
......
..
...
....
......
.
..
......
..
.....
..
..
......
...
......
.
.....
.
..
......
....
.
.
.
.
.
.
......
..
......
.
.....
......
..
...... ..
.......... ..
........
....... ..
.
...
..
...
.
....
.
....

A Tx Range

Tx Range

Tx Range

Tx Range

Figure 1.2: Hidden and Exposed Node S enario


Stations A and C be ome hidden to ea h other as station B
an listen both to A and C, but stations A and C an not listen to ea h other. If

Hidden node Problem:

1.5 Thesis Obje tive and S ope

a pa ket is being transmitted from station A to station B and station C de ides to


start a transmission (being hidden, it does not know of A-B transmission), there
will be ollision at station B.
If station B is sending data to station A, then station C
be omes exposed to B and is for ed to be silent even if it an send data to another
station. This is be ause station C nds arrier busy during transmission of station
B.

Exposed node Problem:

A simple and elegant solution to the hidden node problem is to use small pa kets
alled RTS (Request to Send) and CTS (Clear to Send) for handshaking before transmission of data pa ket. This solution was proposed by Karn [7 in his MACA proto ol
for AX.25.

1.5 Thesis Obje tive and S ope


This work aims at improving the throughput in entralized ooperative wireless ad-ho
networks by use of stations equipped with a new MAC alled dual MAC at one-hop
boundary of the entral oordinating station. The main fo us of this work is to present
the design and ar hite ture of dual MAC. The implementation of Dual MAC in publi ly
available Network Simulator NS-2 is presented, and simulation results are dis ussed.
The s ope of this work is limited to entralized multihop networks to a large extent.
The e e t of node mobility is not onsidered.

1.6 Thesis Outline


The rest of the thesis is outlined in this se tion. In Se ond hapter, di erent medium a ess approa hes, and the hoi es for MACs available for wireless networks are dis ussed.
The operation of IEEE 802.11 MAC is dis ussed in detail. Third hapter dis usses the
operation of IEEE 802.11 MAC in multihop networks, and des ribes the problems of
using the DCF and PCF in su h s enario. In hapter four, we dis uss our approa h to
improve the throughput of multihop ad-ho networks, namely the design and ar hite ture of dual MAC. The hapter ve dis ussed the implementation of dual MAC in NS.
The simulation set up and results are dis ussed in hapter six. Finally, thesis ends with
the on lusion and the s ope for the future work in hapter seven.

Chapter 2
MAC in Multi-hop Ad-Ho
Networks

2.1 Channel A ess Me hanisms


The most important part of a MAC proto ol is Channel A ess Me hanism. The hannel
a ess me hanism is way of regulating the use of physi al hannel among the stations
present in the network. It spe i es when a station an send or re eive data on the
hannel. In this se tion, we dis uss three hannel a ess me hanisms, namely TDMA,
CSMA/CA, and Polling.
2.1.1

TDMA

Time Division Multiple A ess (TDMA) is one of the simplest hannel a ess me hanisms. In this method ea h station sends the data for a nite duration alled time slot.
Typi ally, ea h time slot has xed duration. In TDMA s heme, there should be a way
to gure out the time slot for the transmission. This is done by having a station alled
Base Station whi h is responsible for assigning the time slots to the stations. A set
of time slots forms a TDMA y le or frame whi h repeats at a regular duration. At
the beginning of the time slot, the Base Station sends the allo ation slots for the rest
of the TDMA y le and the stations send the Data in their orresponding time slots.
Typi ally, the stations transmit on one frequen y alled uplink frequen y and re eive on
another frequen y alled downlink frequen y.

Figure 2.1: TDMA hannel A ess Me hanism

2.1 Channel A ess Me hanisms

TDMA suits very well for telephone appli ation be ause of very predi table tra
requirement and is used in ellular telephone networks. However, it does not suite the
pa ket based appli ations where the data tra is bursty and unpredi table. This is
be ause TDMA is very stri t and in exible.
2.1.2

CSMA/CA

CSMA/CA (Carrier Sense Multiple A ess) is derived from CSMA/CD (Collision Dete tion) whi h is the hannel a ess me hanism used in wired Ethernets. Sin e the
transmission range of wireless stations is limited, ollision an not be dete ted dire tly.
This proto ols tries to avoid the ollision. On arrival of a data pa ket from LLC, a
station senses the hannel before transmission and if found idle, starts transmission. If
another transmission is going on, the station waits for the length of urrent transmission,
and starts ontention (i.e. waits for a random amount of time). Sin e the ontention is
a random time, ea h station get statisti ally equal han e to win the ontention.

Figure 2.2: CSMA hannel A ess Me hanism


CSMA/CA is asyn hronous me hanism for medium a ess and does not provide any
bandwidth guarantee. Its a best e ort servi e and is suited for pa ketized appli ations
like TCP/IP. It adapts quite well to the variable tra onditions and is quite robust
against interferen e.
2.1.3

Polling MAC

Polling me hanism in wireless networks is a mix of TDMA and CSMA/CA a ess methods. The base station has total ontrol over medium, but the frame size an be variable.
The base station polls the stations in the network, and the stations reply with the data,
if available. The base station an poll all the stations one by one or follow some intelligent reservation strategy to poll only those stations whi h have data to send. Polling
is very mu h like TDMA, ex ept that it is more exible and allows variable length data
pa kets.
TDMA and Polling methods have disadvantage of being entralized in nature, and
therefore, are not suitable for multihop ad ho networks. This is due to the fa t that
all the stations are not in transmission range of ea h other. CSMA/CA is distributed
in nature, and hen e, is suitable for multihop ad-ho networks. However, it must be
noted that a distributed medium a ess me hanism like CSMA/CD su ers from the
disadvantage of ausing ollision under high load ondition, i.e., if too many stations try
to a ess the medium at the same time, han es of ollision in rease. One way to avoid
6

2.2 Available MACs

Figure 2.3: Polling Me hanism


this disadvantage is to intermix the Polling and CSMA/CA me hanism, as provided in
802.11 PCF mode of medium a ess.
The most suitable a ess me hanism for ad ho networks is the CSMA/CA as it
is distributed and simple in nature. Another reason for using CSMA/CA in wireless
MACs is that wireless LAN ards with CSMA/CA are readily available.

2.2 Available MACs


2.2.1

IEEE 802.11 MAC

The IEEE 802.11 MAC was designed by a ommittee of IEEE. The goal of the ommittee
was to reate a MAC standard for wireless LANs. In the 802.11 ommittee, ea h vendor
has pushed its own te hnology and spe i ations. The result is that the standard is
very versatile and well designed in luding all the optimizations and lever te hniques
developed by di erent vendors. The standard spe i es one MAC proto ol and three
physi al standards: Frequen y Hopping 1Mb/s (Only), Dire t Sequen e 1 and 2 Mb/s
and di use infrared. Subsequently, the standard has been extended to support 2 Mb/s
for Frequen y Hopping and 5.5 and 11 Mb/s for Dire t Sequen e (802.11b). The MAC
has two main modes of operation, a distributed mode (CSMA/CA), and a oordinated
mode. 802.11 also uses MAC level retransmissions, RTS/CTS and fragmentation. The
standard also has optional power management features and optional authenti ation and
en ryption (using the WEP, Wired Equivalent Priva y). Now the standard has also
been extended to be used in 5GHz band (IEEE 802.11a) and some extensions have been
added in 2.4GHz band (IEEE 802.11b) to in rease the bandwidth.
2.2.2

HiperLan-I

The HiperLan standard has been developed by resear hers at ETSI (European Tele ommuni ations Standards Institute). It is developed without any strong vendor in uen e
and is quite di erent from existing produ ts. The standard is quite simple, and uses
some advan ed features. It works in dedi ated bandwidth (5.1GHz to 5.3GHz) and so
does not have to use spread spe trum. The signalling rate is 23.5Mb/s, and 5 xed
hannels are de ned. The proto ol uses a variant of CSMA/CA based on pa ket time
to live and priority, and MAC level retransmissions. The proto ol in ludes optional
en ryption (no algorithm mandated) and power saving.
The ni est feature of HiperLan (apart from the high speed) is the ad-ho routing: if
your destination is out of rea h, intermediate nodes will automati ally forward it through
the optimal route within the HiperLan network (the routes are regularly automati ally
7

2.3 IEEE 802.11 Operation

re al ulated). HiperLan is also totally ad-ho , requiring no on guration and no entral


ontroller.
The main de ien y of HiperLan standard is that it doesn't t provide real iso hronous
servi es (but omes quite lose with time to live and priority), doesn't t fully spe ify the
a ess point me hanisms and hasn't really been proved to work on a large s ale in the
real world. Overhead tends also to be quite large (really big pa ket headers).
2.2.3

HiperLan-II

HiperLan II was the rst standard to be based on OFDM modulation. Ea h sub- arrier
may be modulated by di erent modulations (and use di erent onvolutional ode, a
sort of Forward Error Corre tion, FEC), whi h allow to o er multiple bit-rates (6, 9, 12,
18, 27 and 36 Mb/s, with optional 54 Mb/s), with likely performan e around 25 Mb/s
bit-rate. The hannel width is 20 MHz in 5MHz band, and in ludes 48 OFDM arriers
used to arry data and 4 additional are used as referen es (pilot arriers - total is 52
arriers, 312.5 kHz spa ing).
HiperLan II is a Wireless ATM system, and the MAC proto ol is a TDMA s heme
entrally oordinated with reservation slots. Ea h slot has a 54 B payload, and the MAC
provide SAR (segmentation and reassembly - fragment large pa kets into 54 B ells)
and ARQ (Automati Request - MAC retransmissions). The s heduler (in the entral
oordinator) is exible and adaptive, with a all admission ontrol, and the ontent of
the TDMA frame hange on a frame basis to a ommodate tra needs. HiperLan II
also de nes power saving and se urity features. HiperLan II is designed to arry ATM
ells, but also IP pa kets, Firewire pa kets (IEEE 1394) and digital voi e (from ellular
phones). The main advantage of HiperLan II is that it an o er better quality of servi e
(low laten y) and di erentiated quality of servi e (guarantee of bandwidth).
The IEEE 802.11 standard is the most widespread standard and has been deployed
in wireless LANs. It is also being used in ad-ho network testbeds, and lot of resear h is
going on adapting 802.11 for ad-ho networks. Besides that, it o ers both infrastru ture
mode and infrastru ture-less mode of operation and both an oexist together without
any modi ations. These fa tors make 802.11 MAC very suitable for ad-ho networks
and this thesis dis usses the ad-ho networks using 802.11 MAC.

2.3 IEEE 802.11 Operation


The IEEE 802.11 MAC o ers two kinds of medium a ess methods, namely Distributed
Coordination Fun tion (DCF), and Point Coordination Fun tion (PCF). DCF is the
basi a ess method in 802.11 and requires no infrastru ture. When wireless stations
are within transmit range of ea h other, they form a Basi Servi e Set (BSS), and an
ommuni ate to ea h other using DCF. If the BSS ontains only two stations, it is alled
Independent Basi Servi e Set (IBSS). Many BSSs may be onne ted by a Distribution
System (DS) to form an Extended Servi e Set (ESS). An a ess point (AP) is the
station that provides a ess to DS servi es. The PCF is built on the top of the DCF,
and is also referred to as infrastru ture mode. It requires a polling station alled Point
Coordinator (PC), whi h a ts as ontrolling station during poll. The PCF onsists of
alternating Contention Free Periods (CFP) and Contention Periods (CP). During CFP,
the PC polls other stations in the medium, and during CP, the a ess method be omes
DCF.

2.3 IEEE 802.11 Operation

Required for contentionfree


services
Used for contention services
and basis for PCF
Point coordination
Function (PCF)
MAC
extent
Distributed Coordination Function (DCF)

Figure 2.4: IEEE 802.11 Ar hite ture

2.3.1

DCF Operation

The DCF is the fundamental a ess method used to support asyn hronous data transfer
on a best e ort basis. The DCF is based on CSMA/CA. The arrier sense is performed at
both the air interfa e, referred to as physi al arrier sensing, and at the MAC sublayer,
referred to as virtual arrier sensing. Physi al arrier sensing dete ts presen e of other
users by analyzing the a tivity in the hannel through the re eived signal strength.
A station performs virtual arrier sense by examining the re eived MPDU (MAC
Proto ol Data Unit) information in the header of RTS, CTS and ACK frames. The
stations in BSS use this information to adjust their Network Allo ation Ve tor (NAV),
whi h indi ates amount of time that must elapse until the urrent transmission is omplete and the hannel an be sampled again for idle status.
SIFS
DIFS

RTS
Source

Destination

Data
SIFS

SIFS

ACK

CTS

DIFS

Other

NAV (RTS)

CW

NAV (CTS)
NAV (Data)
Defer Access

Backoff Started

Figure 2.5: DCF a ess using RTS/CTS


Priority a ess to the medium is ontrolled through the use of mandatory interframe
spa e (IFS) time intervals between the transmission of frames. Three IFS intervals are
spe i ed in the standard: Short IFS (SIFS), PCF-IFS (PIFS), and DCF-ISF (DISF).
The SIFS is the smallest and the DIFS is the largest. The SIFS has the highest priority
9

2.3 IEEE 802.11 Operation

to a ess ommuni ation medium. For basi a ess method (without RTS/CTS), the
station waits for DIFS period and samples the hannel again. If the hannel is still idle
the station transmits MPDU. If the re eiving station re eives the pa ket orre tly, it
sends an ACK after waiting for the SIFS time. If RTS/CTS is used then the station
sends a RTS pa ket before sending MPDU. On re eiving RTS, the re eiving station
sends a CTS frame after SIFS time. On re eiving CTS, the sender waits for SIFS time
and transmits the MPDU. Again, the ACK follows after SIFS period. It is also possible
to send the multiple fragments of a MPDU, where ACK to ea h fragment is sent after
SIFS time, and next fragment is sent after SIFS time of ACK.
The ollision avoidan e portion of CSMA/CA is performed through a random ba ko
pro edure. If a station initially senses the hannel busy; then the station waits until
the hannel be omes idle for DIFS period, and then omputes a random ba ko time
within a range alled ba ko window. For IEEE 802.11, time is slotted in time periods
that orresponds to a Slot Time. After ea h unsu essful attempt, the ba ko window
is in reased exponentially until a maximum value. The idle period after the DIFS period
is alled ontention window (CW). The advantage of this hannel a ess method is that
it promotes fairness among stations, but its weakness is that it an not provide delay
bound servi e to the stations.
2.3.2

PCF Operation

The 802.11 MAC o ers ontention free servi e by means of PCF. PCF is optional apability and provides ontention-free (CF) frame transfers. The PCF relies on the point
oordinator (PC) to poll other stations. The polled stations an send the data without
ontending for the medium. In a wireless LAN, the fun tion of a PC is performed by
AP within ea h BSS.
CFP Repitition Interval

CFP

CP

PCF

DCF

CFP Repitition Interval

NAV

CFP

CP

PCF

DCF

NAV

Figure 2.6: PCF a ess


The PCF is required to oexist with the DCF and logi ally sits on the top of DCF
(see Figure 2.4). The PCF onsists of alternating ontention free period (CFP), and
ontention period (CP) as shown in gure 2.6. In the CFP, the PC polls ea h of the
stations present in the BSS. The PC spe i es the start of the CFP by sending a bea on
that ontains the length of CFP duration, among other things. All the stations in the
BSS set their NAV for the duration of the CFP. The PC terminates the CFP by sending
a CF-End frame, and may also terminate it before the advertised CFP duration. The
time di eren e between two bea ons is alled bea on period (BP) or CFP repetition
interval, and is a multiple of bea on frame. The bea on also helps in syn hronization
10

2.3 IEEE 802.11 Operation

and timing. The limits on durations of ea h of the frames are des ribed in the IEEE
802.11 standard [5. In the CP, the stations use DCF to a ess the medium.

Figure 2.7: PCF Transmissions


In CFP, the PC polls ea h station by sending either Poll, Data + Poll or Data +
ACK (for previous frames) + Poll. The polled station sends the data or Data + ACK
(if polled by Data+Poll frame) to PC in response to the PC's poll. It is also possible
to have station to station transmission during the CFP. The PCF o ers onne tion
oriented polled servi e and therefore it redu es ontention to a large extent. It is ideal
for wireless LAN environments where all the stations are onne ted to the AP over a
single hop. However, at low tra the overheads are quit high.
2.3.3

Summary

The DCF is suitable for the ad-ho ontention based servi e, whereas the PCF o ers
polling based servi e. The DCF an exist independently in a network, but the PCF
oexists with the DCF. The DCF an be easily deployed in ad-ho networks, as it does
not require any infrastru ture (like AP in ase of PCF). The PCF an provide onne tion
oriented servi e but is suitable largely for the one-hop wireless LAN setup. However,
simply using PCF in entral part of the multihop network is not enough as the stations
in the PCF mode do not use RTS/CTS ex hange during CFP. The la k of RTS/CTS
ex hange results in hidden and exposed node problems. We dis uss this s enario in
greater detail in next hapter, and propose the solution to the problems en ountered by
simple DCF/PCF operation of IEEE 802.11 MAC in multihop networks.

11

Chapter 3
IEEE 802.11 MAC in Multi-hop
S enario

3.1 Multihop S enario


The IEEE 802.11 MAC is designed for wireless LANs. The requirements of multihop
ad-ho networks are more hallenging than those of wireless LANs. In this hapter, we
investigate the operation of IEEE 802.11 MAC in entralized multihop ad-ho networks.
The terms station and node are used inter hangeably throughout the thesis. Multihop
ooperative wireless ad-ho networks will be simply referred to as multihop networks.
t
t
t
t

t
t

...... ....... ....... ...... ......


......
.....
.....
.....
......
.....
.........
.
.
.
.
.
.
.. ...
..............
.
.........
....
..........
....
.
.
.
.
.
..
.
..
.... ....
...
...
.
.
.
..
..
..
..
.
.
.
.
.
.
.
.... .... ...
.
.
.
.
.
.
..
..
.
.
.... ....
.... ....
...
..
.
.
....
....... .....
..
..
..
....
........
.
.
....
.
.
..
..
.........
.
...
..
.
.
.
.
.
.........
.
.
.
.
.
.
.
...
.... ...
...
.
..
..
..
..
.....
.... ....
.
..
..
..
......
...........
.
.
.
.
....
.. ..
.........
.
.
....
....
.
.....
...............
.....
.... ....
.........
.
.
.
....
.
.
.....
...
.. ..
..... .
..... ....... ....... ...... ....

t
t

Tx Range

t
t
t
t

Central Station

Inner Stations
Boundary Stations
t Outer Stations

.....
.... ....

.......
....
.........

Figure 3.1: Multihop S enario


Consider a multihop entralized s enario, as shown in the gure 3.1. For onvenien e,
the stations inside the network are lassi ed into following ategories:
is the entral ontrolling station. Most of the tra in the network
is dire ted towards it.

Central station

Inner stations

are within one hop boundary of the entral station.

are at one hop boundary of the entral station. These stations


a t as relaying stations for the stations outside the rea h of entral node.

Boundary stations

Outer stations

are outside the ommuni ation range of entral node.


12

3.2 IEEE 802.11 Operation in Multihop Networks

3.2 IEEE 802.11 Operation in Multihop Networks


The 802.11 MAC with DCF mode of operation is the simplest hoi e in multihop adho networks. The reason for the hoi e of DCF is that it does not require any prior
infrastru ture. Two or more stations an ome together and form an BSS. This nature
of DCF is very suitable for ad-ho networks as the ad-ho networks are simply formed
by as set of stations oming together. In this se tion we dis uss the operation of 802.11
MAC in multihop networks, espe ially entralized multihop ad-ho networks
In a entralized multihop network, as shown in Figure 3.1, the node density in
entral region is higher than in the outer region. Most of the tra is dire ted toward
the entral node and boundary stations a t as relaying stations. Therefore, the tra
near the entral station and its one hop neighbours is very high. Sin e the DCF is a
ontention based distributed proto ol, it performs badly in high load onditions. The
poor performan e of DCF is due to fa t that the ollisions in rease as more and more
stations try to a ess the medium at the same time. It is well known that the polling
MAC performs better than pure CSMA/CA under high load onditions. Therefore,
ontention an be de reased by using polling MAC where entral station a ts as polling
station.

..........
..... ...
..........
........
..... .....
...........

........
..... ....
..........

..
..
..
...
..
...
..
..
.
....

..
...

...
..
..
...
...
..
..
..
..
...

..
....

...
...

.
.....

.
.....

...... ........ ........ ....... ....


... .
....... .
......
......
......
......

..........
.... ...
.........

.....
.

u
u

u
...
..

....
..

.....
.

....
..

...
..

.....
.

......
......
......
.......
.......
....... ....
... ........ ........ .......

.
.....

u......
..
....

..........
.
... ....
.........
...
.
..
...
...
..
...
.
..
.
.
.
.
.
.
..
.
..

u
u

....
..

....
..

........
..... ....
..........

...........
.
.... ....
........
.....
.

..
....

.
...

............
..
.... ....
.....

...........
... ..
..........

..
...
..
...
..
..

...........
..
.
... ...
.......
...........
.
.... ....
.......

Central station
u Stations using PCF
.........
.
.... ....
.........
Stations using DCF

Figure 3.2: Hybrid PCF-DCF Operation


The most suitable hoi e for the polling MAC would be PCF mode of 802.11, as it
is an extension of the DCF mode. Ebert et. all [8 have shown that the PCF mode
performs better than DCF when the number of stations in WLAN ell is very high.
Therefore, we make the entral node as Point Coordinator (PC), and it polls all the
inner and boundary nodes during CFP period. This di ers from onventional PCF
operation in WLANs where PC resides within AP. The outer stations still perform DCF
sin e the tra in those regions is not high. The outer stations an send their data
in ontention period (CP) as all the stations perform DCF during CP. We refer this
ombination of PCF and DCF as hybrid operation as shown in gure 3.2.
The hybrid operation seems to be an ideal hoi e in multihop networks, but it gives
rise to following problems:
 The stations that are polled by the Point Coordinator (PC) keep their NAV set

13

3.3 Problem Des ription

during the CFP period, and therefore, an not re eive from outer stations. It an
also be said that the boundary nodes be ome exposed to PC.
 Outer stations be ome hidden to PC, and vi e versa, as there is no RTS/CTS

ex hange between PC and its one hop neighbours during CFP period.

The Figure 3.3 shows that there is no throughput gain with hybrid mode as ompared
with pure DCF mode. In fa t, hybrid mode performs worse than DCF espe ially under
low load.
Comparision of performance of DCF and PCF
1
dcf 10 pkts/s
dcf 20 pkts/s
dcf 30 pkts/s
pcf 10 pkts/s
pcf 20 pkts/s
pcf 30 pkts/s

Packet Delivery Ratio

0.8

0.6

0.4

0.2

0
10

15

20

25
30
CBR connections

35

40

45

Figure 3.3: Comparison of DCF and PCF in Multihop Networks


Due to the above problems, PCF at enter does not in rease throughput, as is
expe ted. Shugong Xu and Tarek Saadawi [9 have also shown that the 802.11 MAC
does not perform well in multihop ad ho networks. This leads to on lusion that
without modi ations, 802.11 MAC with DCF mode or with PCF at enter is not very
suitable for multihop ad ho networks.

3.3 Problem Des ription


Simple DCF is not suitable for entralized multihop network due to ollisions at high
tra . A polling MAC (PCF) is required at the enter of the network to handle high
tra and redu e ollision, but it gives rise to hidden and exposed node problems. The
solution to both of these problems is provided by introdu ing dual Nodes at the boundary
of the entral node. The details of the dual node and its ar hite ture are dis ussed in
next hapter.

14

Chapter 4
Dual MAC

4.1 Need for Dual MAC


The entralized multihop ad ho networks, as explained in earlier hapters, require a
MAC that an provide distributed a ess me hanism. Furthermore, the MAC should
be able to handle high tra in entral region. The pure DCF operation, or hybrid
operation (see Fig. 3.2) do not perform e iently in multihop networks. The reasons
for this ine ien y, as explained in se tion 3.2, are:
1. The DCF does not work well in high load s enario.
2. In ase of hybrid operation, the polling and NAV setting in PCF nodes ause
exposed and hidden node problems, thereby de rease the throughput.
To improve the throughput, boundary nodes should be able to re eive data from
outer nodes during the CFP period (NAV is set). For this the MAC should be able to
re eive even if its NAV is set. Also, transmissions from outer stations should not ollide
with that of PC at boundary stations. To address above problems, we propose to equip
boundary stations with dual MAC.

4.2 Overview of Dual MAC


A dual node is a station whi h has two independent MACs ea h ommuni ating on
di erent logi al hannels. The two MACs are en apsulated inside the dual MAC. The
logi al hannels ould be FDMA or CDMA. Consider the boundary stations in Figures
3.1 and 3.2 that are equipped with dual MACs. One of the MACs uses the PCF and
is termed as PCF MAC. The se ond MAC uses the DCF and is termed as DCF MAC.
The PCF MAC ommuni ates with the PC, and the DCF ma ommuni ates with the
outer nodes. The exposed and hidden node problems in entral region (see se tion 3.2)
are eliminated as follows:
 Boundary stations use the PCF and the DCF on di erent hannels. Therefore,

the transmission of outer node does not ollide with that of PC, and vi e versa.

 The DCF MAC in the dual node an re eive from outer nodes even when the

NAV of PCF MAC is set during CFP period, thereby eliminating exposed node
problem.
15

4.3 Ar hite ture of Dual MAC

4.3 Ar hite ture of Dual MAC


Link Layer

6
?

Dual MAC
DCF
MAC

PCF
MAC

6
?

6
?

Phy

Phy

6
?

6
?

Wireless Channel
Figure 4.1: Ar hite ture of Dual Ma
The dual MAC onsist of two MACs in a single station ea h apable to send and
re eive pa kets independently. Ea h MAC is designed to operate on di erent logi al
hannels. As mentioned earlier, the hannels an be either FDMA or CDMA but it
does not make di eren e as far as the design of the dual MAC is on erned. As shown
in Figure 4.1, the dual MAC is below the Link Layer and basi ally en apsulate two
a tual 802.11 MACs. These two MACs whi h we refer to as PCF MAC and DCF MAC
respe tively, ommuni ate on two di erent logi al hannels. The DCF MAC talks to
the stations that are operating in DCF mode and the PCF MAC talks to the PC and
other stations whi h are in PCF mode. The dual MAC is a layer on the top of these
MACs and single point of interfa e to the Link Layer (LL). For all down-going pa kets
the dual MAC layer a ts as arbitrator and sends on either of the MAC below it. For
the up-going pa kets, the job of dual MAC is simply to hand over the pa ket to the link
layer.

4.4 Operation
A pa ket arriving from link layer is re eived by the dual MAC and handed over to the
MAC at appropriate frequen y. The link layer nds out the MAC address of the next hop
destination by using ARP and hands out the pa ket to the dual MAC layer along with
the destination MAC address. In ase dual MAC, the dual MAC also needs to know the
hannel of the destination station. This ould be done either by ARP table maintaining
information about the hannel on whi h the destination stations is ommuni ation, or
by maintaining a lo al list of stations on ea h hannel. The dual MAC gures out the
hannel of the destination MAC and sends out the pa ket to the appropriate MAC. The
broad ast pa kets like route dis overy pa kets and ARP pa kets are sent to both the
16

4.5 Design Considerations


For outgoing pa kets:
if ( pa ket is broad ast ) {
send pa ket to both PCF MAC and DCF MAC;
} else {
hannel = get_destination_MAC_ hannel();
if ( hannel == DCF_ hannel )
send pa ket on DCF MAC;
else
send pa ket on PCF MAC;
}
For in oming pa kets:
hand over the pa ket to Link Layer;

Figure 4.2: Operation of Dual MAC


MACs. On re eiving a pa kets from lower layer, the dual MAC simply hands it out to
the link layer. The operation of the dual MAC is summarized in gure 4.2

4.5 Design Considerations


The dual MAC is essentially a MAC whi h an a ess two separate logi al hannels.
This an be done in either half duplex way or in full duplex way. In half duplex method,
the logi al hannel (radio frequen y in ase of FDMA, PN ode in ase of CDMA) is
hanged by the radio for a di erent hannel to be a essed. In full duplex method,
both the hannel are a essed simultaneously. This requires that two di erent radios be
employed for full duplex operation. We hoose the full duplex operation sin e it o ers
parallelism of hannel a ess whi h is essential for the boundary nodes. This requires
extra ost, but the addition in ost is justi ed onsidering that very few stations need
to be equipped with the dual MAC.
In the next hapter, we des ribe the implementation of dual MAC in the publi
domain Network Simulatior NS-2 in greter detail.

17

Chapter 5
Dual MAC Implementation in NS

5.1 Existing Node Ar hite ture in NS


target

 uptarget

LL  arptable - ARP

downtarget

IFQ
downtarget

uptarget
- MAC
6uptarget
downtarget
?

ma

Radio
Propagation  propagation
Model

NetIF

hannel ?

6uptarget

Channel

Figure 5.1: Ar hite ture of NS Node


The Figure 5.1 shows the ar hite ture of a NS mobilenode below the Link Layer.
The outgoing pa kets after being pro essed by the routing layer, are handed over to the
link layer (LL obje t) through the target interfa e. The link layer is onne ted to
the ARP module (ARP obje t) through the interfa e arptable . After pro essing the
pa ket and resolving MAC address through ARP, the link layer hands over the pa ket
to the Interfa e Queue (IFQ obje t) through the downtarget interfa e. The interfa e
18

5.2 Node Ar hite ture of Dual MAC

queue is onne ted to the MAC layer (MAC obje t) through interfa e downtarget , and
the pa kets are pulled o by the MAC when required. The link layer also ontains a
referen e to MAC through the ma interfa e. After the MAC has a quired the medium,
it sends it to the physi al layer (NetIF obje t) through the interfa e downtarget . The
physi al layer sends this pa ket over the hannel (Channel obje t) through the hannel
interfa e and the pa ket rea hes to the physi al layer of the station at the other end of
the link.
The outgoing pa ket is handed over by Channel to the physi al layer through the
uptarget interfa e. The physi al layer determines the re eived power levels through
propagation model (propagation ) and pro esses the pa ket. The pa ket after pro essing, is handed over to the MAC layer through uptarget interfa e. The MAC layer
pro esses the pa ket and handles over to the link layer through the interfa e uptarget .

5.2 Node Ar hite ture of Dual MAC


target

 uptarget

LL

downtarget

arptable

ARP

IFQ
downtarget

uptarget

Dual Ma
ma

DCF
MAC

PCF
MAC

downtarget

Radio
Propagation propagation
Model

uptarget

NetIF

hannel ?

6uptarget

Channel

Figure 5.2: Ar hite ture of Dual Node


The gure 5.2 shows the ar hite ture of the dual Node. The dual Node di ers from
the ar hite ture of the original NS node obje t only at the MAC layer. The MAC layer
onsists of a dual MAC obje t whi h en apsulates two 802.11 MAC obje ts within it.
19

5.3 Implementation in NS

All the up oming and downgoing pa kets are rst re eived by the dual MAC. In ase
of downgoing pa kets, the dual MAC determines the hannel on whi h this pa ket is to
be sent, and handles it over to appropriate MAC. In ase of ingoing pa kets, it simply
hands over the pa kets to the MAC orresponding to re eived hannel where it is handed
to link layer after pro essing.

5.3 Implementation in NS
Dual_Ma :: re v(pa ket)
{
hannel_id = hannel id in pa ket header;
if ( re eived pa ket is bea on ) {
p f_ hannel_id = hannel_id;
PC = address of sender;
}
if ( dire tion == DOWN AND pa ket is broad ast) {
send pa ket on DCF MAC;
send pa ket on PCF MAC;
return;
}
if ( dire tion == DOWN AND destination == PC ) {
send pa ket on PCF MAC;
return;
}
if ( dire tion == UP AND hannel_id == p f_ hannel_id ) {
send pa ket on PCF MAC;
return;
}

send pa ket on DCF MAC;


return;

Figure 5.3: Dual MAC implementation in NS


The pseudo ode for working of dual MAC in NS is shown in Figure 5.3. The dual
MAC ( lass Ma Dual) is implemented by deriving the existing Ma lass in NS. The
Ma Dual obje t en apsulates two a tual Ma 802 11 obje ts. The Ma 802 11 lass
and Pa ket lass have also hanged to in orporate dual MAC fun tionality. The Ma
lass has been hange to in orporate the features ne essary for the dual MAC. The
logi al hannels are implemented by in orporating a variable hannel id in the pa ket
20

5.3 Implementation in NS

header. The undesired pa kets (pa kets of di erent hannel) are ltered by examining
this variable in the re eived pa ket header. For the ltering, the hanges have been
done to Ma 802 11 lass. Ea h outgoing pa ket from MAC has its hannel id variable
set to appropriate hannel. All interfa es like uptarget , downtarget , et . have been
hanged su h that all the pa kets that are re eived either from the Physi al Layer or
Link Layer are rst re eived by the dual MAC.
As pointed out in se tion 4.4, the determination of hannel for an outgoing pa ket
an be either done by hanging ARP module or by using a lo al list of stations on
ea h hannel. Currently the se ond approa h is being used. The dual MAC stores the
hannel-id of the PC in a variable as all the tra for PCF MAC goes only to the
PC. All the pa kets that are destined for PC are sent on PCF MAC. For in oming
pa kets, the hannel-id is determined from the pa ket header and pa ket is handed over
to appropriate MAC. The orresponding MAC updates its state and after pro essing
the pa ket and hands it over to link layer.

21

Chapter 6
Simulation and Results

6.1 Simulation Setup


This hapter des ribes the simulation of dual MAC. The simulation are done by using
the publi domain simulator NS-2. The following assumption are made in the simulation:
 The e e t of propagation delay on the model are negle ted. This is fairly re-

alisti onsidering the fa t the area in whi h stations are present is limited to
1500mx1000m and inter-node distan e is of the order of few hundred feet.

 The e e t of hannel errors is ignored in the simulations.


 No stations are operating in power save mode.

A nite bu er is maintained at ea h station. If the bu er lls, the newly generated


pa kets are simply dropped. The safe distan e upto whi h a stations an re eive is
maximum 250m. The interferen e range is 500m. All the pa kets in the DCF mode are
sent using RTS/CTS ex hange. We use onstant bit rate (CBR) tra with data pa ket
size of 512 bytes. The routing proto ol used is DSDV. The reason for hoosing DSDV
proto ol for routing is that it provides onstant routing overhead in ase of stati and
less mobile networks.

6.2 Simple S enario and Results


In order to gain an understating of the dual MAC, we set up a simple s enario as shown
in Figure 6.1. The number of stations in the topology is 18. The re eiving stations for
all the transmitting stations is the station labeled 0. The stations numbered 1, 2, 3, 4,
5, and 6 are inner nodes. These stations are within one hop distan e of the station 0.
The stations 7, 8, and 9 are the boundary nodes. Rest of the stations in the network
are the outer stations.
The Figure 6.2 shows the performan e of the dual MAC as ompared to the DCF
MAC. The number of onne tions is 15, and the pa ket rate is varied to in rease the
load on the network. It an be observed that as the load on the network in reases, the
throughput of network with dual nodes be omes better than that of network with DCF
nodes only. The graph of pa ket delivery ratio for with respe t to o ered load is shown
in Figure 6.3.

22

6.2 Simple S enario and Results

......
..... ....
.......

.
.....
.....
.....
........
.
.....
.... ....
.....
.
......
.
.
.
.....
.....
.....
.
.
.
....
.
.
.
.
..
.....
.... ..
.....
... ..
.....
..... .......
.....
........
.....
....
.....
.
.
.
.
.
...
...........
........
.
.......
.... ....
..... ...
.....
..
.....
.....
.
.
...
.
.
.....
.
.
.
.
.
..
.....
.
.
.
.
.
.
.
.....
....
........
.....
..... .....
.
...........
.....
.......
......
.....
..
.....
.
..... ......
.
.
.
.
...
....
.... ....
.
.....
.............
.......
.
.....
.............
.
.
.
.
.
.............
.....
..............
.
.
.
.
.
.
.
.
.
.
............
....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
....
.. ..
..
.............
.........
.............
.............
..
.............
.............
..
.............
.............
............................
..
.
.
.
.
.
.
.
.
.
.
...................
.
.....
........
........
...........
....
..... .....
.
.
.
.
.... ....
.
..........
............
.......
......
.
...........................
..
.
...................
.
............
.
.............
.
............
.
.............
.............
..
............
..
.............
.............
...............
............
............
.
.............
.....
..
............
..... .....
.............
.......
............
..
............
..
.............
........
............
.
.
.
.
..
... ..
.
.......
.
.
.
.
...... .....
.
........
.........
.
.
... ..
..
......
.
...
..
...
.
....
..........
.
... .....
..........
.
....
.....
.... .....
.
........

12

13

14

.......
.... ....
.......

dire tion of date tra

15

10

Tx Range = 250m ............

18

19

11

...
.
..
..
..
..
..
..
..
..
..
..
..
..
...
.
...
.
...
.
..
.
..
.
......... ..
.
.
... ... ..
......
.
..
.
..
..
..
.
.
..
..
.
..
.
.
.
.
.
.
..
.
.
.
.
.
..
..
.
..
.
........
..
.
.... ....
..
......
..
..
.
...
.
...
.
...
..
..
.
...
.
...
.
..
...
..
..
.
.
.

16

17

Figure 6.1: Simple S enario

Performance of DCF and Dual MAC in simple scenario


160

140

Throughput (Kbytes/sec)

120

100

80

60

40
dual
dcf
20
20

40

60

80

100

120

140

160

180

200

220

Offered Load (Kbytes/sec)

Figure 6.2: Throughput omparision of Dual MAC Vs DCF MAC in simple s enario

23

6.3 Generi S enario and Result

Performance of DCF and Dual MAC in simple scenario


1
dual
dcf
0.95
0.9

Packet Delivery Ratio

0.85
0.8
0.75
0.7
0.65
0.6
0.55
0.5
0

10

15

20

25

30

35

40

Offered Load (Kbytes/sec)

Figure 6.3: Pa ket Delivery Ratio for Dual MAC Vs DCF MAC in simple s enario

6.3 Generi S enario and Result


In this se tion we des ribe the simulation results with random node pla ement. The
stations are pla ed in an area of size 1500m x 1500m. The enter station is pla ed
at the oordinate (750m, 750m). There are 8 boundary nodes pla ed symmetri ally
around the entral station ea h at the distan e of 200m from the entral station. The
number of inner nodes in the topology is 12, and all other stations are more than one
hop away from the entral stations. The maximum number of hops is six. The stations
are randomly pla ed in the area.
The graphs for the \pa ket delivery ratio Vs number of onne tions" for di erent
pa ket rates are shown in the Figures 6.4, 6.5, and 6.6. The x-axis indi ates the number
of CBR onne tions and the y-axis indi ates the ratio of pa kets delivered to the destinations to the number of pa kets sent. The dual MAC o ers substantially higher pa ket
delivery ratio for all the ases. It should be noted that the tra at 30 pa kets/se ond
with more than 10 onne tions orresponds roughly to a load greater than 30x10x512x8
= 1228800 bites/se (1.2Mbps). Even at this load the pa ket delivery ratio of dual MAC
is nearly 84% as ompared to DCF at approximately 50%. The Figure 6.7 shows the
throughput of network with dual MAC and with PCF MAC. It an be learly shown
that at high loads the performan e of dual MAC is substantially better than the DCF
MAC.

24

6.3 Generi S enario and Result

Performance of DCF and Dual MAC at 10 packets/sec


1
dual
dcf

0.6

0.4

0.2

0
10

15

20

25

30

35

40

45

Number of CBR connections

Figure 6.4: Dual MAC Vs DCF MAC at 10 pa kets/se

Performance of DCF and Dual MAC at 20 packets/sec


1
dual
dcf

0.8

Packet Delivery Ratio

Packet Delivery Ratio

0.8

0.6

0.4

0.2

0
10

15

20

25

30

35

40

Number of CBR connections

Figure 6.5: Dual MAC Vs DCF MAC at 20 pa kets/se

25

45

6.3 Generi S enario and Result

performance of DCF and Dual MAC at 30 packets/sec


1
dual
dcf

Packet Delivery Ratio

0.8

0.6

0.4

0.2

0
10

15

20

25

30

35

40

45

CBR connections

Figure 6.6: Dual MAC Vs DCF MAC at 30 pa kets/se

Performance of DCF and Dual MAC


50000
dual
dcf
45000

Throughput (Kbytes)

40000
35000
30000
25000
20000
15000
10000
5000
0

20000

40000

60000

80000

100000

120000

140000

Offered Load (Kbytes)

Figure 6.7: Dual MAC Vs DCF MAC - overall Throughput

26

160000

6.4 Dis ussion on Results

6.4 Dis ussion on Results


As seen from the graphs in the Figure 6.7, the performan e of dual MAC is onsiderably
better than that of DCF MAC. The in rease in the performan e is attributed to following
reasons:
 The PCF and DCF operate at di erent hannels. This means that there is paral-

lelism in the pa ket transmissions. This also eliminates the hidden node problem
in the entralized s enario.

 The dual MAC o ers parallelism by allowing the transmissions and re eptions by

DCF MAC even if the NAV of PCF MAC is set during the poll by PC. This
eliminates the exposed node problem at boundary nodes whi h are exposed to the
PC when the PC polls the stations.

We an see from the graphs that the throughput performan e in rease with dual
MAC is more than twi e that of DCF, whi h is remarkable onsidering that only few
stations have dual MAC.

27

Chapter 7
Con lusion and Future Work

7.1 Con lusion


The design of a MAC that meets the demand of a multihop wireless network is great
hallenge. The restri tions like limited bandwidth, low power, and limited transmission
range make this hallenge even greater. Further, the hidden and exposed node problems
o er even more di ulties by in reasing the han e of ollision. In this work, we have
investigated the usefulness of IEEE 802.11 MAC proto ol using the PCF and DCF
me hanisms. We nd that without modi ations, the PCF and DCF are not very useful
in multihop networks. We also nd that the high tra in the entral region of a
real-life entralized multihop network makes the DCF unsuitable. We investigated the
use of PCF in entral region to ope up the high tra requirements in the entral
region. However, it gave rise to exposed and hidden node problems. The dual MAC
was designed to eliminate exposed and hidden node problems in the entral region of a
entralized multihop network. The thesis dis ussed the design and ar hite ture of dual
MAC along with its implementation in NS.
The results show that the dual MAC performs reasonably better than the DCF a ess
me hanism. However, the dual MAC requires two physi al radios and two separate
hannels { one bound to PCF and another to DCF. We may however, note that the
throughput gain is worth the ost of dual MACs sin e only few nodes need to be equipped
with dual MAC.
The main fo us of this thesis is to suggest a modi ation to the existing IEEE
802.11 MAC so as to make it suitable in multihop ad-ho networks, espe ially in the
real life entralized networks. The results presented in the thesis are appli able for stati
s enario, nevertheless, the dual MAC is expe ted to perform better even under mobile
s enario. The e e t of mobility on performan e of dual MAC still remains to be seen.
One of the reasons we ould not study the e e t of mobility was la k of implementation of asso iation and disasso iation of stations to PC in PCF in NS. La k of this
feature restri ted to allow stations to move and dynami ally asso iate and disasso iate
from the PC.

7.2 Related Work


The dual MAC aims at eliminating the exposed and hidden node problems at the enter
of the topology where PCF is used. Very little work has been done on entralized
topologies, and using PCF in an ad-ho networks. Most of the previous work has been
28

7.3 Future Work

on entrated on DCF mode of 802.11. Poojary, Krishnamurthy and Dao [10 have tried
hanging power of ontrol pa kets (RTS, CTS and ACK) with stations having power
ontrol apabilities. They show that it does not in rease the throughput of the network
due to overheads. Deng and Haas propose dual Busy Tone Multiple A ess (DBTMA)
[11 for eliminating the exposed and hidden node problems. DBTMA used di erent
hannels for data and ontrol pa kets. Nasipuri and Das [12, [13 have proposed MAC
with multiple hannels, and shown throughput improvement.
The dual MAC has been designed keeping in the mind the number of hannels
available in 802.11. The most ommon method of physi al a ess in available WLAN
ards is the DS-CDMA. The 802.11 with DS-CDMA allows three di erent hannels. The
dual MAC an use any two of them. Another advantage of using dual MAC is that very
few stations need the dual MAC apability (Boundary Nodes) and rest of the stations
an ontinue using existing WLAN ards. However, the dual MAC requires that PCF
and DCF be done at di erent frequen ies.

7.3 Future Work


One very useful extension to this work will be a s enario where one of the stations
automati ally be omes the PC after seeing high tra in its vi inity. This requires a
me hanism for a node to be ome PC and inform other station about the same. This
will mean that the high tra regions perform s heduled MAC, i.e., the PCF, and the
surrounding stations use the DCF. In this s enario, the dual nodes promise to be of
great use by providing parallelism at the boundary stations. The dual MAC an also
be enhan ed to use both the MACs in DCF and/or PCF mode, even if the station is in
a region where only PCF or DCF is being used. (Currently only possible operation is
that of one of MACs using the DCF and another using the PCF at boundary stations.)
This enhan ement will mean that the parallelism o ered by two hannels will in rease
the throughput onsiderably.

29

Bibliography
[1 C. Perkins, Network Working Group, RFC 3220: IP Mobility Support for IPv4,
http://www.ietf.org/rf /rf 3220.txt.
[2 Charles E. Perkins et. all, Mobile Ad Ho Networking Working Group,
Ad ho On-Demand Distan e Ve tor (AODV) Routing, http://www.ietf.org/
internet-drafts/draft-ietf-manet-aodv-12.txt.
[3 David B. Johnson, David A. Maltz, Yih-Chun Hu, and Jorjeta G. Jet heva,
IETF MANET Working Group, The Dynami Sour e Routing Proto ol for
Mobile Ad Ho Networks (DSR), http://www.ietf.org/internet-drafts/
draft-ietf-manet-dsr-07.txt.
[4 Charles E. Perkins, Pravin Bhagwat, Highly dynami Destination-Sequen ed
Distan e-Ve tor routing (DSDV) for mobile omputers, ACM SIGCOMM Computer
Communi ation Review, v.24 n.4, p.234-244, O t. 1994.
[5 IEEE Std. 802.11, Wireless LAN Media A ess Control (MAC) and Physi al Layer
(PHY) Spe i ations, 1999.
[6 Brian P. Crowe, I. Widjaja, J. Kim, P. Sakai, IEEE 802.11 Wireless Lo al Area
Networks, IEEE Communi ations Magazine, September 1997.
[7 Phil Karn, MACA - A New Channel A ess Method for Pa ket Radio, ARRL/CRRL
Amature Radio 9th Computer Networking Group paper 801.22/92-39, Mar h, 1992.
[8 Andreas Kopsel, Jean-Pierre Ebert, and Adam Wolisz, A Performan e Comparision
of Point and Distributed Coordination Fun tion of an IEEE 802.11 WLAN in the
presen e of Real-Time Requirements, Pro . of 7th Intl. Workshop on Mobile Multimedia Communi ations (MoMuC2000), O tober 23-26, 2002.
[9 Shugong Xu, Tarek Saadawi Does IEEE 802.11 MAC Proto ol Work Well in Multihop Wireless Ad Ho Networks?, IEEE Communi ations Magazine, p.130-137, June
2001.
[10 Neeraj Poojary, Srikanth V. Krishnamurthy, and Son Dao, Medium A ess Control
in a Network of Ad Ho Mobile Nodes with Heterogeneous Power Capabilities, IEEE
International Conferen e on Communi ations (ICC 2001), volume 3, p.872-877, 2001.
[11 J. Deng, and Z. J. Haas, Dual Busy Tone Multiple A ess (DBTMA): A New
Medium A ess Control for Pa ket Radio Networks, IEEE ICUPC'98, Floren e, Italy,
O tober 5-9, 1998.

30

BIBLIOGRAPHY
[12 A. Nasipuri, and S. R. Das. A Multi hannel CSMA proto ol for multi-hop wireless networks, Pro . of IEEE Wireless Communi ations and Networking Conferen e
(WCNC'99), September 1999.
[13 Asis Nasipuri, and Samir R. Das, Multi hannel CSMA with Signal Power-Based
Channel Sele tion for Multihop Wireless Networks, Pro eedings of the IEEE Fall
Vehi ular Te hnology Conferen e (VTC 2000), September 2000.
[14 The VINT proje t, NS notes and do umentation, editors: Kevin Fall and Kannan Varadhan, http://www.isi.edu/nsnam/ns.
[15 Ad Kamerman and Leo Monteban, WaveLAN-II: A High-Performan e Wireless
LAN for the Unli ensed band, 1997.

31

You might also like