A selective border-casting zone routing protocol for ad-hoc networks

Abstract

One of well-known routing protocol for ad hoc networks is zone routing protocol (ZRP). The performance of ZRP is better than other protocols. However, many useless control packets are used resulting in the increase of network load and decrease of network performance. In this paper, we propose a selective border-casting zone routing protocol (SBZRP) to reduce the network load by limiting the number of control packets when the protocol searches for a new route. The performance evaluation via simulations shows that the SBZRP has better performance than ZRP.

Full text

Page 1

A Selective Border-casting Zone Routing Protocol for Ad-hoc Networks
Leonard Barolli*, Yoshitaka Honma**, Akio Koyama***, Arjan Durresi****, Junpei Arai**
*Department of Information and Communication Engineering
Fukuoka Institute of Technology (FIT), Japan
barolli@fit.ac.jp
**Graduate School of Science and Engineering, Yamagata University, Japan
g03209@dipfr.dip.yz.yamagata-u.ac.jp, j_arai@astro.yamagata-cit.ac.jp
***Department of Informatics, Yamagata University, Japan
akoyama@yz.yamagata-u.ac.jp
****Department of Computer Science, Louisiana State University, USA
durresi@csc.lsu.edu
Abstract
One of well-known routing protocol for ad-hoc networks is
Zone Routing Protocol (ZRP). The performance of ZRP is
better than other protocols. However, many useless control
packets are used resulting in the increase of network load
and decrease of network performance. In this paper, we
propose a Selective Border-casting Zone Routing Protocol
(SBZRP) to reduce the network load by limiting the number
of control packets when the protocol searches for a new
route. The performance evaluation via simulations shows
that the SBZRP has better performance than ZRP.
1. Introduction
With the development of computer industry, the computer
size is getting smaller but with rich functionality. New types
of computers such as note Personal Computer (PC), Personal
Digital Assistant (PDA), the increase of network speed and
decrease of transmission cost have increased the number of
users and computers resulting in very fast growing of
Internet. But now the users want to connect to the network at
any place and any time. The wireless mobile networks and
devices are becoming increasingly popular to provide users
the access anytime and anywhere. In order to connect mobile
terminals to the network generally are used wireless LANs
[1]. But, in wireless LAN, the communication between
mobile terminals is done using access points and thus the
movement of mobile terminals is limited. Presently, to deal
with this problem ad-hoc networks are proposed. The Mobile
Ad Hoc Networks (MANETs) [2] do not use any fixed
infrastructure. The nodes of MANET intercommunicate
through single-hop and multi-hop paths in a peer-to-peer
fashion. Intermediate nodes between two pairs of
communication nodes act as routers. Thus the nodes operate
both as hosts and routers. The nodes are mobile, so the
topology of the network may change rapidly and
unexpectedly.
Much work has been done on routing in MANETs [3,4,5].
Many protocols and algorithms such as Destination-
Sequenced Distance-Vector (DSDV) protocol, cluster-based
routing algorithms, Dynamic Source Routing (DSR) protocol,
Ad hoc On-demand Distance-Vector (AODV) protocol, Zone
Routing Protocol (ZRP), Temporally Ordered Routing
Algorithm (TORA), and Associative Bit Routing (ABR)
have been proposed. Among these protocols, the ZRP has a
wide application [6]. However, when the protocol searches
for a new route, it sends many useless control packets, which
increase the network load and decrease the network
performance.
In this paper, we propose a Selective Border-casting Zone
Routing Protocol (SBZRP) to reduce the network load by
limiting the number of control packets when the protocol
searches for a new route. The performance evaluation via
simulations shows that the SBZRP has a good behavior and
better performance than ZRP.
This paper is organized as follows. In Section 2, we
introduce ad-hoc network routing protocols. In Section 3, we
explain ZRP. In Section 4, we present the proposed SBZRP.
In Section 5, we discuss the performance evaluation. Finally,
some conclusions and future work are given in Section 6.
2. Ad-hoc network routing protocols
The effectiveness of a routing protocol in ad-hoc
increases as network topology information becomes more
detailed and up-dated. Also, the topology may change quite
often, requiring large and frequent exchanges of data among
network nodes.
Existing ad-hoc routing protocols can be classified into
two groups: proactive and reactive routing protocols.
Proactive routing protocols attempt to continuously
evaluate the routes within the network, so that when a packet
needs to be forwarded, the route is already known and can be
immediately used. Proactive protocols can be divided in
Distance Vector (DV) protocols (e.g. DSVD) and Link State
(LS) protocols (e.g. Optimized Link State Routing (OLSR)).
Proceedings of the 15th International Workshop on Database and Expert Systems Applications (DEXA’04)
1529-4188/04 $ 20.00 IEEE

Page 2

The LS protocols converge faster than DV protocols, but at
the expense of significantly more control traffic. Motivation
to both improve protocol convergence and to reduce control
traffic has led to the development of proactive path finding
algorithms by combining the DV and LS protocols. Path
finding algorithms like Wireless Routing Protocol (WRP) [7]
are able to eliminate the “counting-to-infinity” problem and
reduce the occurrence of temporary loops.
Reactive protocols invoke a route determination
procedure on an on-demand basis. The reactive route
discovery is usually based on a query-reply exchange, where
the route query is flooded through the network to reach the
desired destination. The on-demand discovery of routes can
result in much less traffic than DV or LS protocols. However,
the reliance on flooding may still lead to considerable control
traffic in the highly versatile ad-hoc networking environment.
The advantage of proactive schemes is that route
information is available when needed, resulting in little delay
prior to data transmission. In contrast, reactive schemes may
produce significant delay in order to determine a route when
route information is needed, but not available.
Routing schemes, whether proactive or reactive, require
some exchange of control traffic. This overload can be quite
large in ad-hoc networks, where the topology frequently
changes. Reactive protocols produce a large amount of traffic
by effectively flooding the entire network with route queries.
Therefore, they can not be used for real-time communication
applications. Pure proactive schemes are likewise not
appropriate for ad-hoc networks, as they continuously use a
large portion of the network capacity to keep the routing
information current.
Proactive protocols tend to distribute topology changes
widely in the network, even though the creation/destruction
of a new link at one end of the network may not be a
significant piece of information at the other and of the
network. Also, since ad-hoc network nodes may move quite
fast, and as the changes may be more frequent than the routes
requests, most of this maintained routing information is
never used. This results in further waste of the network
capacity.
3. ZRP
The comparison of proactive and reactive schemes shows
that what is needed is a protocol that initiates the route-
destination procedure on-demand, but at limited search cost.
The ZRP is a hybrid reactive/proactive scheme. On one hand,
it limits the scope of the proactive procedure only to the
node’s local neighborhood. On the other hand, the search
throughout the network, although global, can be performed
efficiently by querying selected nodes in the network, as
opposed to querying all network nodes.
In ZRP, a node proactively maintains routes destinations
within a local neighborhood, which is considered as a routing
zone. A node routing zone is defined as a collection of nodes
whose minimum distance hop from the node is no greater
than a parameter referred to as the zone radius. Each node
maintains its own routing zone, but the routing zones of
neighborhood nodes overlap. For construction of a routing
zone the information of neighbor nodes is needed. A
neighbor node is defined another node that a direct
communication can be established and is one hop away. In
ZRP, the IntrAzone Routing Protocol (IARP) is used for
routing within a zone and IntErzone Routing Protocol (IERP)
for routing beyond the routing zone.
The neighbor discovery information is used for proactive
monitoring of routing zones through IARP. The IARP is a
simple timer-based LS routing protocol. To track the
topology of “i” hop routing zone, each node periodically
broadcast its link state for a depth of “i” hops, which is
controlled by a Time-To-Live (TTL) field in the update
messages. The nodes after receiving the IARP packets
transmit them to the neighbor nodes until the Hop Counter
(HC) number becomes the same as zone radius. However, if
the HC in the IARP packet is bigger than the HC recorded in
a node routing table, it is considered that packets are coming
from the routing loops. This procedure is carried out
periodically for all nodes and the routing tables are updated.
When a request comes from nodes inside the zone radius, the
routing table is checked and the packets are sent immediately
to the destination node.
The IERP is responsible for routes located beyond the
routing zone. IERP uses a query-response mechanism to
discover routes on demand. The IERP is distinguished from
standard flooding algorithms by exploiting the structure of
the routing zone, through a process known as border-casting.
Border-casting is a packet delivery service that allows a node
to efficiently send a message to its peripheral nodes. The
ZRP provides this service through a component called
Border-cast Resolution Protocol (BRP).
An IERP route query is triggered at the network layer,
when a data packet is destined for a node that does not lie
within its routing zone. The source generates a route query
packet, which is uniquely identified by a combination of the
source node’s ID and request number. The query is then
border-casted to all the source’s peripheral nodes. Upon
receipt of a route query packet, a node adds its ID to the
query. The sequence of recorded node ID’s specifies an
accumulated route from the source to the current routing
zone. If the destination does not appear in the node’s routing
zone, the node border-casts the query to its peripheral nodes.
If the destination is a member of the routing zone, a route
reply is sent back to the source, along the path specified
reversing the accumulated route. As with standard flooding
algorithms, a node will discard any replicated route query
that it has previously encountered.
4. Proposed Protocol
The proposed SBZRP uses for intra-zone routing the IARP
the same as ZRP, but uses a new IERP for inter-zone routing.
To explain IARP let consider Fig.1. The node S generates
the IARP packet (S is Source Node (SN) and the HC is 1)
and sends it to all neighbor nodes (nodes A, B and C). The
node after receiving IARP packet updates its own routing
table using IARP packet information. The neighbor nodes of
node S (A, B and C) send the received IARP packet to their
neighbor nodes (see Fig. 2). The nodes are moving so the
route information may be inappropriate after a period of time.
For this reason, TTL parameter is used as shown in Table 1.
Proceedings of the 15th International Workshop on Database and Expert Systems Applications (DEXA’04)
1529-4188/04 $ 20.00 IEEE

Page 3

When the HC in the IARP packet is bigger than the HC in
the routing table, the packets are discarded (see solid line
arrows in Fig.2).
Let explain IERP by using Fig.3. Let consider that SN S
received a request to find a route to Destination Node (DN) J.
If there is not routing information in the routing table, an
IERP packet is generated. The IERP packet has this
information: SN = S, DN = J and the Border-cast Hop
Number (BHN) = 1. The format of IERP packet is shown in
Fig.6(a). The generated IERP packet is sent to all nodes
which are in the zone border (see Fig.4), which have the
same HC with zone radius. Therefore, there are many nodes
within the zone radius. The IERP request format after node S
has sent the request is shown in Fig.6 (b).
The nodes in the zone boundary which receive the
ReQuest Packet (RQP) increase the BHN by 1. Then, the
node checks its routing table to find a route to the DN. If the
node does not find the DN in its table, the border-cast
procedure continues (see solid line arrows in Fig.4).
However, the IERP RQ packet is not sent to the SN (see
dotted line arrows in Fig.4). In the case when a node find a
DN or is itself the DN, the node sends back to the SN (S) the
IERP packet (see Fig.5). This packet is called Reply Packet
(RP). In this example, in the routing table of node G is
recorded the information of node J, because node J is a
member of routing zone G. Therefore, node G sends backs
the IERP RP to S by reversing the accumulated route.
The reply route format in this case is shown in Fig.6 (c).
After receiving the IERP RP, the SN knows the route to send
the data packets. The data packet format is shown in Fig.7.
When a SN finds a route to the DN, in ZRP for inter-zone
routing, the IERP RQPs are border-casted to all zone nodes.
But, when a route if found, for a short period of time may be
this route is still a good one and can be used for routing
without searching a new route. In SBZRP, when a new
search is carried out for the same node, the number of IERP
packet sending directions is limited. The IERP RQPs are
saved for awhile in a buffer and if there are requests for
routing in the same DN, the IERP RQP are sent only to the
nodes of the previous search as shown in Fig.8.
When a node moves but is inside the zone as shown in
Fig.9, the intermediate nodes of the previous recorded path
are C and G. In SBZRP, when the IERP RQP arrives at node
C, node J is inside node C zone, thus node C sends the IERP
RP to SN S. When a node moves outside a zone, but the
period of time from the last search is short, it can be
considered that node is not too far from the route recorded in
the IERP. Therefore, a node in the IERP can search to find a
new route to node J. In Fig. 10, when a RP arrives in node G,
but it has not found a route to node J, a new search is started
from node G. Thus, the number of the border-cast nodes and
IERP packets can be decreased resulting in the increase of
the throughput and the decrease of packet mean delay.
5. Performance Evaluation
We consider the following scenarios for simulations. In
Scenario I, the number of nodes is 10, the field size is 500 m
x 500 m, the transmission distance is 100 m, and the nodes
are not moving (network topology is not changed).
Fig. 1. IARP packet broadcasting.
Fig. 2. IARP packet forwarding.
Table 1. Routing table of node H in Fig. 2.
DN
S
C
HC
2
1
SN
C
C
TTL (ms)
1500
2000
Fig. 3. IERP border-cast.
Fig. 4. IERP packet transmission.
S
G
B
I
E
D
C
H
A
S
G
B
I
E
D
H
A
C
S
A
C
B
E
D
H
F
G
I
J
S
A
C
B
E
D
H
F
G
J
I
Proceedings of the 15th International Workshop on Database and Expert Systems Applications (DEXA’04)
1529-4188/04 $ 20.00 IEEE

Page 4

Fig. 5. IERP RP transmission.
DN SN Intermediate BN Packet Type BHN
(a) IERP packet format.
DN SN Intermediate BN Packet Type BHN
J S Null RQP 1
(b) IERP request packet format of Fig. 3.
DN SN Intermediate BN Packet Type BHN
J S C, G RP 2
(c) IERP reply packet format of Fig. 5.
Fig. 6. IERP packet format.
DN SN Intermediate BN Packet Type Data
J S C, G Data Data
Fig. 7. Data packet format after Fig. 5 operation.
Fig. 8. Route search of SBZRP.
In Scenario II, the number of nodes and the field size is
the same as Scenario I, but the generation rate is considered
30 ms (fix generation rate) and the nodes are moving. As
moving model for Scenario II, we consider Random
Waypoint Model (RWM) and the moving speed is 10 m/s.
Until the network becomes stable is needed a period of time.
Fig. 9. A case when DN moves inside the zone.
Fig. 10. A case when DN moves outside the zone.
For this reason, first 10 seconds are not considered for
simulation calculation. The simulation continues until the
number of generated packets reaches 1000. We compare the
performance of the proposed SBZRP with ZRP for the
number of packets sent to the DN without loss, the
throughput, packet mean delay, and link usability.
In Fig. 11 is shown the characteristic of packet arrival
rate versus the generation rate for Scenario I. The results
show that when the data generation time is small (the
network load is high), the SBZRP has better performance
than ZRP. This is because when the network load is high
many packets collide and the network performance degrades.
In Fig. 12 is shown the characteristics of throughput
versus packet generation rate for Scenario I. In this figure,
the same as in Fig.11, when the network load is high, the
SBZRP has better behavior than ZRP. When the network
load is low, the throughput is decreased for both protocols.
This is because the number of generated packets is low
which results in low throughputs. Also, when the network
load is high, the throughput is decreased. The reason is that
the number of collided packets is increased, which results in
decrease of network throughput.
The mean delay versus packet generation rate
characteristic for Scenario I is shown in Fig.13. When the
network load is high, the mean delay of ZRP is higher than
SBZRP. This is because when the ZRP searches for a new
route, the number of IERP RQP is increased, thus the node
buffer is congested, which results in the increase of mean
delay.
S
A
C
B
E
D
H
F
G
I
J
S
A
C
B
E
D
H
F
G
I
J
S
A
C
B
E
D
H
F
G
I
J
I
S
A
C
B
E
D
H
F
G
J
Proceedings of the 15th International Workshop on Database and Expert Systems Applications (DEXA’04)
1529-4188/04 $ 20.00 IEEE

Page 5

The link usability versus node stop time for Scenario II is
shown in Fig. 14. By using RWP, a node is moving for a
period of time then stops for a moment of time. This pattern
is repeated in a random way. When the node stops for a short
period of time, that means the moving degree is high, the
SBZRP has higher link usability than ZRP. When node
moving degree is high, the route search failure becomes high.
For the SBZRP, if the route search fails, a new route search
starts from the failed node. Thus, the new route search time
is shorter than ZRP and the number of data sent to the DN
becomes high.
From the simulation results, we conclude that when the
network load and node moving degree is high, the
performance of SBZRP is better than ZRP.
6. Conclusions
In this paper we proposed a new zone routing protocol for
ad-hoc networks called SBZRP. The performance of the
proposed protocol was evaluated by computer simulations
using two Scenarios: Scenario I (nodes were not moving) and
Scenario II (nodes were moving). From the simulation results,
we conclude as follow. In Scenario I, when the network load
is high, the number of arrived packets to DN without loss of
SBZRP is higher than ZRP, resulting in better throughput of
SBZRP. Also, the mean delay of SBZRP is lower than ZRP.
For Scenario II, when the node moving degree is high, the
SBZRP has high link usability than ZRP.
In this research, we used a two way transmission link.
However, in real wireless environments also exists one way
transmission links. Therefore, we would like to consider in
the future also this kind of environment. The SBZRP is
implemented in Java and embedded in mobile cellular
terminals. But during simulations, we had some problems
with terminal memory. For this reason, we would like to
implement the protocol in C language to deal with this
problem. We would like to carry out extensive simulations
for different number of nodes and node moving degrees.
References
[1] IEEE802.11, “Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) Specifications”, First
Edition, 1999.
[2] MANET,http://www.ietf.org/html.charters/manet-charter.
html
[3] C.E.Perkins, Ad Hoc Networking, Addison Wesley, 2001.
[4] E.M. Royer and C.K. Toh, “A Review of Current Routing
Protocols for Ad Hoc Mobile Networks”, IEEE Personal
Communications, Vol.6, No.2, 1999, pp. 46-55.
[5] K.C. Toh, Wireless ATM and Ad-Hoc Networks, Kluwer
Academic Publishers, 1997.
[6] Z.J. Haas and M.R. Pearlman, “The Performance of
Query Control Schemes for the Zone Routing Protocol”,
IEEE/ACM Transactions on Networking, Vol.9, No.4,
2001, pp. 427-438.
[7] S. Murthy and J.J. Garcia-Luna-Aceves, “An Efficient
Routing Protocol for Wireless Networks”, MONET, Vol.
1, No.2, 1996, pp. 183-197.
0
100
200
300
400
500
1015 203040 50 100 500
Packet Generation (ms)
Fig. 12. Throughput vs packet generation.
ZRP
SBZRP
Throughputs (Kbps)
0
100
200
300
400
500
600
1015 2030 40 50 100 500
Packet Generation (ms)
Fig. 13. Mean delay vs packet generation.
ZRP
SBZRP
Mean Delay (ms)
0
5
10
15
20
25
012369 12 15 18 20
Node Failure Time (s)
Fig. 14. Link usability vs node stop time.
ZRP
SBZRP
Link Usability (%)
0
20
40
60
80
100
10 15
Packet Generation (ms)
20 30 4050 100 500
ZRP
SBZRP
Arrival Rate (%)
Fig. 11. Packet arrival rate vs packet generation.
Proceedings of the 15th International Workshop on Database and Expert Systems Applications (DEXA’04)
1529-4188/04 $ 20.00 IEEE