Source Routing Based on Destination Frequency Analysis for Wireless Mobile Ad Hoc Networks

safflowerpepperoniΚινητά – Ασύρματες Τεχνολογίες

24 Νοε 2013 (πριν από 3 χρόνια και 11 μήνες)

78 εμφανίσεις





Source Routing Based on Destination
Frequency Analysis for Wireless Mobile
Ad Hoc Networks


Susan Rea & Dirk Pesch

Adaptive Wireless Systems Group

Department of Electronic Engineering

Cork Institute of Technology, Cork, Ireland

Tel. +353 21 4326100 Fax:
+353 21 4326625

Email: {srea, dpesch@cit.ie}



Abstract


A mobile ad hoc network (MANET) consists of a collection of mobile hosts
(or nodes) that can form an arbitrary network topology. As a consequence of
such an environment it is necessary that these mob
ile nodes also provide
routing functionality in the propagation of information throughout the network.
This paper proposes a hybrid approach to MANET routing that is source
initiated and involves maintaining a destination path cache at each node, based
on
a frequency analysis of previous destination route discoveries. A
performance evaluation of the proposed approach is also presented.



1. Introduction


The proliferation of wireless portable devices as part of everyday life, such as mobile
phones, laptops
and PDAs, is leading to the possibility for spontaneous or ad hoc wireless
communication. With these types of devices there is a fundamental ability to share
information. However, communication over existing infrastructures may be precluded due
to deficien
t facilities, or impractical in terms of time, expense and power.

Networks suited to this type of random communication are known as Mobile Ad
-
hoc
NETworks (MANETs), where a MANET is a self
-
organising collection of wireless
mobile nodes that form a tempora
ry network without the aid of a fixed networking
infrastructure or centralised administration [1], as shown in Figure 1. Node mobility
causes frequent and unpredictable changes in this arbitrary network topology and as such,
MANET routes have to cope with
frequent changes and may consist of many hops
through other network hosts. As a consequence of this dynamic topology the design of
efficient routing protocols is an exigent challenge and crucial problem.



Figure
1
. MANET Architect
ure

In ad hoc networks messages sent by a node may be received simultaneously by all
nodes within its transmission range, i.e. by its neighbours. Messages requiring a
destination outside this local neighbourhood zone must be hopped or forwarded by these
ne
ighbours, which act as routers, to the appropriate target address. As a consequence of
node mobility fixed source/destination paths cannot be maintained for the lifetime of the
network. As a result of this, a number of routing protocols have been proposed
and
developed for wireless ad hoc networks [2,3,4,5,6]. These protocols have been derived
from distance vector and link state techniques and involve determining the shortest path to
a destination in terms of distance or link cost. Such protocols are classi
fied as proactive,
reactive or hybrid, depending on how route maintenance and discovery is performed.


2. Overview of Routing Techniques


A routing protocol must determine the most efficient path between source and
destination nodes so that packets may be
sent throughout a network. A routing protocol
must address the following issues:



Scalability


ability to support a large number of nodes & networks



Ability to adapt to a varying topology & traffic levels in terms of speed and
efficiency



Route discovery



Ro
ute maintenance

Routing protocols can be classified in three categories:



Centralised or Distributed



Adaptive or Static



Reactive or Proactive or Hybrid

A centralised protocol involves all decisions being made at a centre node whereas a
distributed one inc
ludes all nodes in the routing decision. With an adaptive protocol route
information can be modified as a consequence of network status, such as traffic levels or
topology changes, whereas static based protocols update route information at periodic
interva
ls. Reactive protocols perform route discovery when needed (on demand protocols)
and proactive protocols undertake route discovery before it is needed using routing tables
that are updated periodically (table driven protocols). A hybrid protocol, such as Z
one


Routing protocol [7], combines the features of both reactive and proactive schemes to
make a more efficient protocol. Standard routing protocols (those used in fixed networks)
are based on either Distance Vector (DV) or Link State (LS) algorithms [3,4]
.

DV based routing protocols, such as ARPANET and RIP [3], are classed as being
decentralised and static and are based on the Distributed Bellman
-
Ford (DBF) algorithm
[8]. LS routing protocols, such as OSPF [3], are based on Dijkstra’s shortest path
algori
thm [8,9]. These protocols involve each node sustaining a complete map of the
network topology and the cost to reach each destination node.

Conventional routing protocols, based on DV and LS techniques, were designed for
static infrastructure based netw
ork topologies and node mobility was not considered. For
mobile networks, unlike rigid networks, periodic transmission of topology information is
required for efficient routing and this wastes battery power [1]. As nodes must both send
and receive this inf
ormation it makes conservation of energy difficult. Also, periodic
transmission wastes network bandwidth [10] as routing information will often not change
from one update to the next. In a static infrastructure based network node links are
considered bi
-
di
rectional and of equal quality but this may not be the case for ad
-
hoc
networks, thus routes determined by standard protocols may be only unidirectional. Some
of these problems have been approached and solved [2] but conventional protocols as such
still re
main unsuitable for ad
-
hoc networks but have been used as a basis for designing
MANET routing protocols.


2.1 Ad
-
hoc Routing Protocols


An ad
-
hoc routing protocol must be distributed as each node should be involved in
route discovery making the routing in
formation and link costs more reliable. With a
wireless environment and mobile nodes all links should be considered as possibly being
unidirectional and a protocol should be able to adapt to this constraint. In terms of battery
consumption a protocol must
be energy efficient as the sending/receiving of routing
information consumes battery power. Also quality of service issues such as time
-
delay and
throughput are factors considered by real
-
time applications. Consequently, the significant
characteristics of

an ad
-
hoc routing protocol are [3,4]:



Dynamic Topology



Restricted Bandwidth



Erratic Capacity Link, possibly unidirectional



Energy Constraints

Based on when and how route discovery is initiated, there are three main classes of
MANET routing protocols [1,
3,4]:



Table Driven (Proactive)


each node maintains a table of all possible paths to
every node within a network.



On Demand (Reactive)


Route discovery is only initiated when there is a need to
establish a communications link between nodes.



Hybrid


t
his is a fusion of proactive and reactive protocol techniques.

The routing protocol presented in this paper is a source rooted hybrid technique that
maintains proactive routes to destination nodes that are frequently required with all other


routes being g
enerated on demand. The advantage of proactive protocols is that when a
route is required it has already been determined, however, sustaining complete network
topological route information requires large protocol overhead. Reactive schemes involve
delays i
n determining routes as route information may not be known at the time of the
route request and these protocols necessitate significant control traffic to determine routes
[3,4]. This hybrid technique combines the advantages of proactive and reactive proto
cols
as it minimises the amount of routing information that is to be maintained and new routes
can be generated on demand.


Discover Network
Topology
Determine Local
Neighborhood
Build Initial
Node Route Caches
Destination Frequency
Analysis
Route Cache
Update
Send Data
Packet
Next hop in
Route Cache
Yes
Transmit
Packet
No
Generate Route
On Demand
Cache Route
Data
Successful
Yes
No
Drop
Packet
Generate
Delay
N
Retries?
No
Yes
Send
Acknowledgement
Node
Movement
Yes

Figure
2
. Protocol Flowchart


3. Source Routing Protocol Based on Destination F
requency Analysis


The proposed protocol, depicted in Figure 2, is based on a hybrid approach towards
packet routing that collaborates to allow route discovery. The protocol is aimed at an
environment with a relatively high density of nodes and short
-
range

communication. Each
node proactively maintains a route cache consisting of path information for the most


frequently accessed destination nodes. A source node S desiring to transmit a data packet
to a destination node D must acquire the next hop node along

the path to D. If this
information is not readily available then route discovery is performed on demand.


3.1 Route Discovery


Initially, when the protocol is instigated route discovery is performed with each node
determining their one hop neighbours (nod
es that are within transmission range of the
source node) and building an adjacency matrix D, describing the network topology, using
the All
-
Pairs Shortest Path algorithm based on the following recurrence:


)
]
,
[
]
,
[
,
]
,
[
min(
]
,
[
1
1
1





k
k
k
k
j
k
D
k
i
D
j
i
D
j
i
D


where D[i,j]
k

is defined to be the length of the shortest path from node i to node j using
only nodes numbered from 1,2,…k as possible transitional hop nodes [9]. D is then used
to determine route cache for each node. Thi
s initial route cache contains a list of possible
destinations, the next hop on the route path to those destinations and a hop count metric
that gives the total number of hops to each destination.

For a set period of time every node records the destinatio
n addresses that it requires for
this phase and also the rate of recurrence for this target node i.e. its frequency. The
longevity of a route cache entry is directly related to its frequency. After some fixed
interval of time a destination frequency analys
is is performed and those destinations with
the highest frequency (above some threshold value) will be sustained in a route cache by
each node. Subsequently, destinations not included in this route cache will be generated
on demand with this path informati
on then being added to the route cache. Route caches
are periodically updated based on past destination frequencies.

After an initial route cache update, based on a previous destination frequency analysis,
if a source S needs to transmit a data packet to s
ome node D that is not a neighbour of S
and S does not have a next hop entry for D in its route cache it performs ensuing route
discoveries as follows: S broadcasts a local Route Request (RREQ) message that is
received by its one hop neighbours. The RREQ m
essage header will distinguish the source
node S and the destination D that is the objective of the route discovery. These neighbour
nodes examine their route caches for an entry for D and document the destination and
source of the RREQ. If some node finds

an entry it returns a Route Reply (RREP)
message to S indicating that it is the next hop and the hop count metric for D. If multiple
RREP messages are received then the node with the lowest hop count metric is selected as
the next hop. Node S then increas
es this hop count by one and records this information in
its route cache.

However, if after time t node S has not received a RREP for D (meaning that it one hop
neighbours have no entry for D) it retries route discovery and retransmits this RREQ to its
lo
cal neighbours. This RREQ is the same as the last RREQ recorded by these nodes and
this indicates that the previous route discovery was unsuccessful and these nodes then
become the new message source and propagate an RREQ for D to their immediate
neighbour
s, which also note this RREQ. If this action is successful in uncovering a next
hop for D then a RREP message is propagated back towards S with intermediate nodes
caching next hop information and hop counts for D. Again, if this action is fruitless the
new

source nodes perform route discovery after time t. After some time t
1

>> t if the


original source node S has not received a RREP for D then it ceases attempting to uncover
a route to D and D is deemed unreachable until the next topology update with all fu
rther
data packets for D being dropped. This information is then propagated throughout the
network using local broadcasts.



3.2 Data Packet Transmission


When transmitting a data packet either directly between a source node and a
destination or over inte
rmediate hop nodes the destination node is responsible for sending
an acknowledgement message to the source node. Failure of the source node in receiving
this acknowledgement results in the packet being considered as undelivered. When
transmitting data pac
ket up to N attempts (N
≥ 1) are possible and if after the N
th

endeavour delivery is still not feasible the packet is dropped. There is a delay time
between consecutive retransmissions that is determined using a modified version of the
Truncated Binary Exponential Back
-
off algori
thm [11] as follows:



For retransmission attempt j (1 ≤ j ≤ N)



Set Maximum delay MD to 2
j
-

1



Generate a random delay k in the range j → MD



Wait for time k to elapse



Retransmit


Repeat if not successful



If j = N then drop packet

The delay time between
transmission retries grows exponentially with the delay time
range being determined as shown in section 4.1.



3.3 Node Mobility


Network topology changes due to node mobility will in effect cause the protocol to be
reinitialised as shown in Figure 2.


4
. Network Implementation


The proposed protocol is implemented in C++ using the Communication Network
Class Library (CNCL) [12]. For network simulation the system is modelled so that the
performance of the system can be expressed in terms of generating and

transmitting
events, with events being likened to data packets. The network is modelled using the
following objects:



Traffic Generator


this generator produces data packet traffic with a Poisson
distribution and uniformly generates a random source no
de and destination node
address over which a packet is to be sent.



Node


this class models the network nodes. All nodes retain the following
information:



Current X,Y grid position





Addresses of one hop neighbouring nodes



Cost associated with links to tho
se neighbours





Current state


ready or busy


Figure 3.

Delay Time Range
Table 1.

Packets Dropped for Delay Range R
i




Route Cache



Destination frequency counter


4.1 Network Simulation


So as to determine o
ptimal suitable values for delay time range and number of
retransmission attempts made before a packet is dropped the network was set up with the
following properties:



Dimension






1200m by 1200m



Number of Nodes



50



Number of Packets



10000

This

simulation topology was based on a static network without frequency updates and
was used to determine the delay time range for packet retransmission. Of three possible
ranges R
i
, the delay range chosen is j → 2
j


1, j = 1..5, was chosen as the optimal on
e
from simulation as shown in fig. 3 and table 1 as this range resulted in the lowest average
number of packets being dropped. The number of retry transmission attempts was set to 5
as values beyond that when compared with the delay range possible did not
significantly
reduce the number of packets dropped to warrant such large delays between retransmitting
packets, using the above delay range, as shown in fig. 4 and table 2.


Figure 4
. Transmission Retry
T
able 2
. Packets Dropped for Retry Attempt

Retry

Attempt j

R1


0...2
j
-
1

R2

0…5

R3

j...2
j
-
1

0

282

282

282

1

319

319

342

2

215

255

124

3

104

192

28

4

61

122

5

5

36

89

17

Average

170

210

133


Delay Time Range
0
100
200
300
400
0
1
2
3
4
5
6
Retry Attempt j
Packets Dropped
R1
R2
R3
Retry Attempts
0
10
20
30
0
5
10
15
20
Number of Retries
Packets
Dropped
Retry

Attempt

Number of


Packets Dropped

5

17

7

27

9

23

11

15

13

11

15

9






Route Cache

Before Update

Destination

Frequency

Counter

Destination

Next Hop

Hop Count Metric



0

9

3

0

0

1

9

2

1

1

2

7

2

2

0

3

9

2

3

0

4

4

0

4

0

5

7

2

5

0

6

9

2

6

0

7

7

1

7

1

8

8

1

8

0

9

9

1

9

0

Table 3
. Destination Frequency Analysis and Route Cache for Node4


Route Cache

After Update

Destination

Next Hop

Hop Count Metric

1

9

2

7

7

1

0

9

3

Table 4.

Reduced Route Cache for Node4 after Update


4.2 Destination Frequency Analysis


During the pr
otocol initialisation phase each network node develops a route cache that
contains next hop information for every node and the total number of hops to each of
those nodes, based on the adjacency matrix D. To begin with the next hop for every node
is proact
ively known. However, after some time t
2

destination frequency analysis is
performed and all route caches are updated maintaining only the most frequently used
next hop data and all other next hop information must be generated on demand until the
next cach
e update. Using the following network topology shown in tables 3 and 4 are the
results of a frequency update for a single node:



Dimension









1200m by 1200m



Number of Nodes






10, addressed as Node0, Node1,…,Node9



Total Number of Packets




2
0, packet counter PC counts from 1 to 20



Frequency Update






5



Hold RC Data







30%

The timing of the frequency updates is implemented in terms of the number of packets to
be sent over the network. For this simulation there were in total 20 pack
ets to be sent and
the frequency update was set to 5 meaning that whenever the following is true:



PC modulo 5 = 0, PC > 0



a frequency up date will be initiated. After the frequency update the top 30% of the route
cache data will be retained so for 10 nodes

the cache entries for the 3 most frequently
required destinations are preserved.


As can been seen from table 3 the destinations with the highest frequencies are 1 and 7.
This destination/frequency counter is sorted in descending order and the top 3 desti
nation
entries (1,7,0) are used to build the updated route cache, as shown in table 4. After a cache
regeneration the destination/frequency counter is reset to 0 and the destination frequencies
are recorded again until the next update.


5. Conclusions and
Further Work


This paper has introduced a hybrid MANET routing protocol that is source initiated
with mobile hosts in a network acting as routers to propagate information with route
information concerning the most frequently accessed destination nodes bein
g cached and
all other destination path data is generated on demand.


Optimal values for the occurrence of frequency updates and the percentage of a route
cache information to be retained after these updates must be determined through
simulation over more
dense networks. Along with this, the aspect of node mobility must
be incorporated into the protocol. With the protocol fully developed its performance must
be evaluated in terms of a comparison with standard ad hoc routing protocols, using
packet delivery
rate, packets dropped, network density and energy consumption as
appraisal parameters.


References


[1]

D. B. Johnson,
Routing in Ad Hoc Networks of Mobile Nodes
, Proceedings of the IEEE Workshop on
Mobile Computing Systems and Applications, December 1994

[2]

C. E. Perkins, P. Bhagwat,
Highly Dynamic Destination

Sequenced Distance
-
Vector Routing (DSDV)
for Mobile Computers
, Proceedings of ACM SIGCOMM’94, September 1994

[3]

C. E. Perkins, Editor,
Ad Hoc Networking
, Addison
-
Wesley, Boston etc., December 2000

[4]

C
-
K. Toh, Ad Hoc Mobile Wireless Networks Protocols and Systems, Prentice Hall, Inc., New Jersey,
December 2001

[5]

A. Boukerche,
Performance Comparison and Analysis of Ad Hoc Routing Algorithms
, Proceedings
IEEE International Conference on Performanc
e, Computers and Communications, 2001

[6]

Internet Engineering Task Force, MANET Working Group,

http://www.ietf.org/html.charters/manet
-
charter.html

[7]

Z. J. Haas, M. R. Pearlman,
The Zone Routing Protocol (ZRP) for Ad Hoc Networks
, Work in Progress,
Int
ernet Draft, MANET Working Group, draft
-
zone
-
routing
-
protocol
-
00.txt, 20 November 1997

[8]

E. Kreyszig,
Advanced Engineering Mathematics
8
th

Edition, John Wiley & Sons, New York etc.,
1999.

[9]

S. S. Skiena,
The Algorithm Design Manual
, Springer
-
Verlag, Ne
w York etc., 1998.

[10]

M. W. Subbarao,
Dynamic Power
-
Conscious Routing for MANETs: An Initial Approach
, Journal of
Research of the National Institute of Standards & Technology, Volume 104, Number 6, Nov.
-
Dec.
1999

[11]

D. Bertsekas, R. Gallagher,
Data Net
works
, 2
nd

Edition, Prentice Hall, NJ 07458,1992.

[12] A. Speetzen, M. Junius, M. Steppler, M. Büter, D. Pesch, and others,
Communication Network Class
Library
, Edition 1.12, September 1999.