Saad Biaz, Bing Qi, Shaoen Wu
Dept. of Computer Science and Software Engineering, Dustan Hall 107, Auburn University, AL 36849, USA
Yiming Ji
Department of Computer Science,University of South Carolina, Hargray Building, Beaufort, SC 29909, USA
Multi-hop network, Dynamic Source Routing, Multi-Radio, Dynamic Source Routing with Multi-radio Ex-
Performance on multi-hop networks suffers from limited throughput capacity and poor scalability problems as
the network size or density increases (Gupta and Kumar, 2000). One way to alleviate these problems is to equip
each node with multiple radios. As hardware cost drops, this approach becomes more and more appealing and
feasible. By tuning radios into non-interfering channels, the wireless spectrum can be more efficiently utilized,
thus enhancing the whole network performance. This work extensively evaluates Dynamic Source Routing
Protocol(DSR) (Johnson and Maltz, 1996) for multi-radio multi-hop networks. Through simulations, DSR
with multi-radio extensions exhibits an overall performance improvement for throughput, packet delivery rate
and latency.
Unlike traditional wireless networks, a multi-hop net-
work is a collection of independent wireless nodes
that establish a network normally without any pre-
established infrastructure. Each node can transmit
data packets to other nodes within its radio range, and
forward packets on behalf of other nodes.
Despite their flexibility and convenience to sup-
port diverse applications (Cordeiro and Agrawal,
2002), multi-hop networks are not yet widely de-
ployed since they perform poorly as the number
of nodes and/or hops increases (Gupta and Kumar,
2000). The poor performance mainly results from
the inability of a wireless radio to transmit and to
receive at the same time. Such a weakness halves
the forwarding node capacity; in addition, simultane-
ous transmissions on the same frequency channel and
the sub-optimal MAC back-off mechanisms exacer-
bate the limited capacity problem.
Many approaches have been proposed to alleviate
the capacity problem. One approach is to explore di-
rectional antennas (Ko et al., 2000; Choudhury and
Vaidya, 2002). Since directional antennas are able
to focus energy in a given direction, they have some
advantages over omni-directional antennas in multi-
hop networks, such as less interference, longer trans-
mission range and the increasing potential for spatial
reuse, etc. Such features result in higher multi-hop
capacity and better connectivity. However, replacing
omni-directional antennas with directional ones will
not fully exploit all the advantages. A number of ad-
justments are required at each layer of the networking
protocol stack to accommodate directional antennas
(e.g. broadcast problem, discovery of neighbors, etc).
Another approach consists of using multiple chan-
nels for each node. For example, the widely
deployed IEEE802.11b standard supports 3 non-
overlapping channels without interference. Users in
infrastructure-based wireless networks already suc-
cessfully exploit the multi-channel feature: differ-
ent access points are assigned to different channels
such that neighboring cells communicate on non-
overlapping channels concurrently. Such channel as-
signment results in lower interferences and higher ca-
pacity (Lee et al., 2002; Tzamaloukas and Garcia-
Luna-Aceves, 2001). However, this method cannot be
applied “as is” to multi-hop wireless networks since
neighboring nodes must communicate on the channel
with same frequency. Some researchers suggest al-
Biaz S., Qi B., Wu S. and Ji Y. (2007).
In Proceedings of the Second International Conference on Wireless Information Networks and Systems, pages 65-69
DOI: 10.5220/0002147300650069
lowing each node to use multiple channels following
a hopping sequence (Tyamaloukas and Garcia-Luna-
Aceves, 2000). Receiver and sender nodes will fol-
low some channel hopping sequence till they lock on
a common channel. Unfortunately, a non-negligible
switching latency (especially for off-the-shelf wire-
less network interfaces (Inc., 2004)) would negatively
impact the performance.
Due to ever decreasing hardware cost, the switch-
ing latency can be eliminated by equipping each node
with multiple radios: the radios on each node would
be tuned on multiple non-overlapping channels, thus
wireless nodes can send and receive independently
and simultaneously on multiple channels. Therefore,
with multiple radios on each node, the available spec-
trum could be more efficiently shared. However, after
adding one or more radios to each node in multi-hop
networks, routing data packets becomes more chal-
lenging in such scenarios. Although most well-known
multi-hop routing algorithms have been so far exten-
sively studied in case of mono radio nodes (Bouk-
erche, 2004; Broch et al., 1998), far fewer studies ex-
ist for evaluating their performance when nodes are
configured with multiple radios (Pirzada et al., 2006).
Some other research works also investigated the
multi-radio nodes effect but most of them focus on the
link-quality routing protocols and new routing metric
such as (R. Draves and Zill, 2004; Couto et al., 2003).
On the contrary, our work only extends Dynamic
Source Routing(DSR) (Johnson and Maltz, 1996) to
take advantage of multi radio feature and evaluates the
Raw performance improvement due to multi-radio in
multi-hop networks.
Extensive simulations are conducted to show that
DSR with multi-radio extensions is very efficient for
high traffic loads and exhibits a high delivery rate as
well as a lower latency.
The rest of the paper is organized as follows: Sec-
tion 2 briefly introduces the basic DSR scheme and
describes our extensions to DSR to take advantage of
the multi-radio nodes. Section 3 outlines the simula-
tion settings. And Section 4 demonstrates and ana-
lyzes the experiment results. Finally, Section 5 con-
cludes the paper.
Routing protocols for multi-hop networks are nor-
mally classified as reactive (or On demand) and
proactive (Royer and Toh, 1999) protocols. Reactive
routing protocols (e.g. AODV (Perkins and Royer,
1999), DSR (Johnson and Maltz, 1996)) create and
Source Request Destination
Figure 1: Simple RREQ Packet Format.
maintain a route between a pair source-destination
only when the source node needs to send packets
to the destination node; In contrast, proactive rout-
ing protocols (e.g. OLSR (Clausen et al., 2003),
STAR (Garcia-Luna-Aceves and Spohn, 1999)) re-
quire wireless nodes maintain routing tables for all
nodes on the network.
Broch et. al (Broch et al., 1998) evaluated multi-
ple ad hoc routing protocols and concluded that Dy-
namic Source Routing (DSR) (Johnson and Maltz,
1996) is one of the best in terms of resources con-
sumption in single radio multi-hop network. While
DSR has been extensively studied in single radio net-
works, there is no work to our knowledge that eval-
uates its performance directly over multi-radio multi-
hop networks. Thus this work extends DSR to work
with multi-radio nodes scenarios and evaluates the
performance of such extensions through extensive
DSR (Johnson and Maltz, 1996) is a reactive rout-
ing protocol specially designed for wireless multi-hop
networks and it is based on the concept of source rout-
ing as the source node specifies in the packet’s header
the sequence of nodes to reach the destination.
The basic idea of DSR routing protocol lies in
its route discovery process. When a source node S
intends to communicate with a destination node D
whose route is unknown, the source node S initializes
a route discovery process by flooding out a route re-
quest packet (RREQ) to all its neighbors (RREQ sim-
ple format is illustrated in Figure 1). On receiving a
RREQ packet, node A checks the destination address
in RREQ packet’s header: if A is the target node, it
returns a route reply packet (RREP) to the initiator
node S by following the path which is typically the
reverse of RREQ Route-Record field. RREP will con-
tain the sequence of nodes on the path from source S
to target D; Otherwise, node A is just one intermediate
node, thus node A caches the RREQ packet, appends
its own address to the RREQ’s Route-Record field,
and rebroadcasts the updated RREQ. Node A discards
the RREQ packet in case the same RREQ packet has
already been previously received. After the source S
receives RREP, it caches the route to send subsequent
data packets to the specific destination node.
When a node is configured with more than one
radio, radio indices are needed to make DSR aware
of the existence of multiple radios. Figure 2 shows
a simple network scenario with four nodes, in which
WINSYS 2007 - International Conference on Wireless Information Networks and Systems
Figure 2: Simple Multi-radio multi-hop network.
Source Request Destination
Radio Index
Figure 3: Modified RREQ Packet Format.
every node has two independent radios (respectively
represented by a triangle and a circle). The radios on
each node are set to different non-overlapping chan-
nels such that they can independently work without
any interferences. Since each node may have one or
more radios in the multi-radio network, the route for
a given path must include the radio index information
to each hop. So on Figure 2, one possible path should
be specified as: [(S,1)-(A,2)-(B,1)-D] where 1 and 2
indicate the radio index on each node.
Therefore some slight changes should be made
to traditional single radio DSR in order to perform
properly in multi-radio scenarios: each node needs to
broadcast RREQ packets on all its radios; when an in-
termediate node gets the RREQ packets from any of
its radios, the node should append not only its own
address, but also the radio index to the route record to
indicate on which radio it recieves the RREQ packet.
As shown on Figure 3, a radio index field is added to
DSR RREQ packet header.
Furthermore every radio within each node needs
to send out a copy of RREQ packet in multi-radio
networks. Thus many more RREQ packets will be
generated causing more RREQ packet collisions than
single radio networks, also at the same time increase
the likelihood to lose RREQ packets with good routes
information (broadcast packets are not retransmitted
when lost). In single radio multi-hop network, DSR
forwards only the first RREQ packet and discards fur-
ther RREQ with the same request id and source ad-
dress. Here, in multi-radio network, in order to in-
crease the chances to get the shortest routes, the nodes
are required to forward RREQ packets more greedily
than single radio networks: That is the intermediate
node forwards again a RREQ packet previously seen
as long as the hop count is lower.
Table 1: Experimental Parameter.
Simulation Time 120 Seconds
Simulation Field 1500m*1500m
Propagation Mode Two-ray Ground
Traffic Mode CBR
Transmission Range 250 m
Number of Connections 5,10,15,20
Packet Size 1000 bytes
Traffic Interval 30 ms
Interface Queue 50
3.1 Experimental Set Up
The efficiency of DSR on a multi-radio network is
evaluated using ns-2 simulator. In our simulations,
all nodes have the same number of radios and the
radios on each node are assigned to different non-
overlapping channels. The same channel allocation
scheme is used for all wireless nodes.
The experimental multi-hop network consists of
50 nodes which are randomly positioned a field of
. All nodes are configured with two radios
tuned to two non-interfering channel. Source node
and destination node are randomly selected to start
UDP connections. Each UDP connection sends CBR
traffic with 1000 byte data packets every 30ms. To
test the effect of various traffic loads, the number of
connections is varied from 5 to 20 taking values 5, 10,
15, 20. For a given set of parameters, the experiment
is repeated for 50 times with different starting time.
The performance metrics are obtained by averag-
ing the results of 50 simulation runs. DSR on single
radio network and DSR on multi-radio network are
evaluated with same simulation setting. The simula-
tion parameters are summarized in Table 1.
3.2 Comparison Metrics
Average Throughput, Packet Delivery Rate, Average
End-to-end Delay, and Average Path Length are the
metrics used to evaluate DSR routing protocol on
multi-radio networks (Broch et al., 1998):
Average Throughput measures the total number of
data bits successfully received during the unit ex-
perimental time.
Packet Delivery Rate measures the number of
end-to-end packets successfully received over the
total number of data packets sent.
DSR with One Radio
DSR with Two Radios
Average Throughput (Mbps)
5 10 15 20
Number of Connections
Figure 4: Average Throughput Vs. Connections.
Average End-to-end Delay measures the average
latency time for all successfully received data
Average Path Length is the average number of
hops a packet took to reach its destination.
The simulations evaluated the impact of traffic load
by varying the number of connections taking values
5, 10, 15, or 20. Promising performance results are
obtained with extended DSR for multi-radio nodes.
All plots have the number of connections on the
x-axis. Figure 4 plots the average throughput respec-
tively achieved by DSR with single radio nodes and
by extended DSR with multi-radio nodes, which il-
lustrates a significant throughput improvement of ex-
tended DSR with the multi-radio nodes over DSR
with single-radio nodes. The average throughout im-
provement reaches up to 93% for different connec-
tions. As the number of connections increases, the
throughput enhancement is more dramatic.
The Packet Delivery Rate is plotted on Figure 5.
As illustrated, more data packets are lost as the
number of connections increases (traffic load in-
creases). Losses may result from unavailable or in-
correct routes, overflow of the buffers, and many other
reasons. The Packet Delivery Rate for extended DSR
on the multi-radio networks is usually better than
DSR on single radio networks regardless of the num-
ber of connections.
Figure 6 plots the Average End-to-end delay re-
spectively for DSR with single radio nodes and for
extended DSR over multi-radio nodes. The delays
significantly vary with the number of connections.
As the number of connections increase, medium con-
DSR with One Radio
DSR with Two Radios
Average Delivery Rate (%)
5 10 15 20
Number of Connections
Figure 5: Average Packet Delivery Rate Vs. Connections.
DSR with Two Radios
DSR with One Radio
Average Delay (seconds)
5 10 15 20
Number of Connections
Figure 6: Average Delay Vs. Connections.
tention and retransmission increase causing more de-
lay at each hop and more retransmissions. Although
the average delay for extended DSR with multi-radio
nodes is quite high for heavier traffic load, it remains
dramatically lower than with DSR on single radio
nodes. On average, the delay is four times smaller.
Figure 7 plots the Average Path Length. The left-
most column indicates the average shortest hop count
that physically existed based on perfect and global
knowledge of the network topology. The middle
column and the rightmost column display the aver-
age hop length respectively taken by extended DSR
with multi-radio nodes and DSR with the single radio
nodes. Results show that extended DSR with multi-
radio nodes finds more often the existing shortest path
especially when the number of connections is high.
WINSYS 2007 - International Conference on Wireless Information Networks and Systems
DSR with Two Radios
DSR with One Radio
Shortest Hop Existed
Average Path Length
5 10 15 20
Number of Connections
Figure 7: Average Path Length Vs. Connections.
This work evaluates the Dynamic Source Routing
protocol performance on a multi-radio multi-hop net-
work through extensive simulations.
Experimental results show a considerable im-
provement of overall performance over DSR with sin-
gle radio nodes in terms of throughput, end-to-end
delay, and packet delivery ratio. Therefore, by sim-
ply enabling Dynamic Source Routing to work with
multi-radio nodes, we can exploit the wireless spec-
trum more efficiently. However, DSR adopts the
shortest hop count to select the routing path, which
is quite limited and does not fully take advantage of
multi-radio feature in wireless multi-hop networks. It
would be of high interest to consider paths with min-
imal interference and higher throughput. The authors
are currently designing a metric similar to the one pro-
posed by Draves and Zill (R. Draves and Zill, 2004)
to take into account multiple criteria such as channel
conditions and radio interferences.
Boukerche, A. (2004). Performance evaluation of routing
protocols for ad hoc wireless networks. Mob. Netw.
Appl., 9(4):333–342.
Broch, J., Maltz, D. A., Johnson, D. B., Hu, Y.-C., and
Jetcheva, J. (1998). A performance comparison of
multi-hop wireless ad hoc network routing protocols.
In Mobile Computing and Networking, pages 85–97.
Choudhury, R. and Vaidya, N. (2002). On ad hoc routing
using directional antennas.
Clausen, T., (editors), P. J., Adjih, C., Laouiti, A., Minet, P.,
Muhlethaler, P., Qayyum, A., and L.Viennot (2003).
Optimized link state routing protocol (olsr). RFC
3626. Network Working Group.
Cordeiro, C. and Agrawal, D. (2002). Mobile ad hoc net-
Couto, D. D., Aguayo, D., Bicket, J., and Morris, R. (2003).
A high-throughput path metric for multi-hop wireless
Garcia-Luna-Aceves, J. J. and Spohn, M. (1999). Source-
tree routing in wireless networks. In ICNP, pages
Gupta, P. and Kumar, P. R. (2000). The capacity of wire-
MATION THEORY, 46:388–404.
Inc., M. I. P. (2004). Max2820, max2820a, max2821,
max2821a 2.4ghz 802.11b zero-if transceivers data
sheet rev. 04/2004. California, USA.
Johnson, D. B. and Maltz, D. A. (1996). Dynamic source
routing in ad hoc wireless networks. In Imielinski
and Korth, editors, Mobile Computing, volume 353.
Kluwer Academic Publishers.
Ko, Y.-B., Shankarkumar, V., and Vaidya, N. H. (2000).
Medium access control protocols using directional an-
tennas in ad hoc networks. In INFOCOM (1), pages
Lee, Y., Kim, K., and Choi, Y. (2002). Optimization of
ap placement and channel assignment in wireless lans.
lcn, 00:0831.
Perkins, C. E. and Royer, E. M. (1999). Ad-hoc on-demand
distance vector routing. wmcsa, 00:90.
Pirzada, A. A., Portmann, M., and Indulska, J. (2006). Eval-
uation of multi-radio extensions to aodv for wireless
mesh networks. In MobiWac ’06: Proceedings of the
international workshop on Mobility management and
wireless access, pages 45–51, New York, NY, USA.
ACM Press.
R. Draves, J. P. and Zill, B. (2004). Routing in multi-radio,
multi-hop wireless mesh networks. In Proceedings of
the 10th annual international conference on Mobile
computing and networking, pages 114 128, Philadel-
phia, PA, USA.
Royer, E. and Toh, C. (1999). A review of current routing
protocols for ad-hoc mobile wireless networks.
Tyamaloukas, A. and Garcia-Luna-Aceves, J. J. (2000).
Channel-hopping multiple access. In ICC (1), pages
Tzamaloukas, A. and Garcia-Luna-Aceves, J. J. (2001).
A receiver-initiated collision-avoidance protocol for
multi-channel networks. In INFOCOM, pages 189–