International Journal of Computational Engineering Research(IJCER)

International Journal Of Computational Engineering Research (ijceronline.com) Vol. 3 Issue. 2

Throughput of Wireless Multi-Mesh Networks: An Experimental Study
P Ramachandran
Department of Electronic Systems Engineering, IISc, Bangalore, India

Abstract
To overcome the capacity and interference problem of Wireless Mesh Network (WMN), we explore the possibility of
increasing the coverage area of WMN by running multiple WMNs in parallel and interconnecting them. Theoretical
achievable capacity from [3] to every node in a random static wireless ad hoc network with ideal routing is estimated as
O(1/√n log n), where n is the total number of nodes in the network. Therefore, with increasing number of nodes in a network,
throughput capacity becomes unacceptably low. In this work we report our throughput measurements that show throughput in
a WMN for different path length is almost the same, with nodes across two WMNs of the same path length. We propose to
interconnect the networks by using multiple wireless adapters in gateway node configured with the SSID of the networks in
operation. To achieve our goal we exploited the DSR protocol feature of assigning locally unique interface indices to its
adapters.

Keywords :MR-LQSR, ETX. PktPair, RTT, ETX, WCETT, ARQ and MCL
1. Introduction
IEEE 802.11 standard provides infrastructure-less operating mode called as ad hoc mode. Absence of infrastructure
and self-configuration capability makes ad hoc networks suitable for low-cost applications, but results in wireless medium
instability and low connectivity. Hence WMN is proposed to increase the ad hoc connectivity because WMN extends the
coverage area of communication by using multi-hop communications. WMN is a communication network made up of radio
nodes organized in a mesh topology, wherein each node in the network may act as a router, regardless of whether it is
connected to another network or not. Nodes allow continuous connections and reconfigurations around the broken or blocked
links/paths by hopping from node to another node until data in the form of packets/frames reach the destination from the
source node.Mesh Routers/Clients and Mesh Gateways are the components of WMN. Mesh clients are often laptops, cell
phones, desktops and other wireless devices. WMN is a special type of wireless ad hoc network where each user node operates
not only as host but also a router. Traffic is forwarded to and from internet connected gateways in a multi-hop fashion. WMN
nodes usually do not have strict power constraint, because it is assumed that all the nodes are almost stationary, but this need
not be so. Networking infrastructure is decentralized and simplified because each node needs only to transmit as far as the next
node. Each node is connected to several other nodes with redundant routes and if one drops out of the network, its neighbors
find another route unlike WLAN access points, which can relay messages on behalf of others. WMNs are dynamically self-
organizing and self-configuring, with the nodes in the network, thus increasing range, stability and the available
bandwidth.The reminder of this paper is organized as section-2 presents the existing technology and motivation to carry out
this work, section-3 describes the methodology of the work, section-4 presents the results and observations, and finally
section-5 concludes our work and future directions.

2. BACKGROUND
Off-the-shelf IEEE 802.11-based wireless interfaces has made easy and inexpensive to setup WMNs with mobile and
PDA devices and be used as a private network. The performance of the WMN is based on the routing protocol and routing
algorithm that is used. Hence research is active in the field of routing protocols and routing metrics to improve the WMN
performance. Most of the routing protocol is evaluated based on simulation. One more issue on which performance of WMN
improve is by using multiple interfaces in a node instead of a single interface. But limitation with using multiple interfaces in a
node are several like, in conventional network for each physical network interface used, should be assigned with different IP
addresses, but a node should have a single unique address irrespective of number of interfaces being used to communicate in
the network. Therefore a suitable protocol with mechanism for communication between nodes equipped with single and
multiple interfaces or mixture of single, multiple interfaces nodes. Routing metric worth should be used, which is directly
responsible for the throughput performance. Board crosses talk, radiation leakage and an adequate distance required to
separate multiple antennas in a node are some of the issues with use of multiple interfaces, hence using multiple interfaces in a
node without taking proper measure would yield less performance in WMN.

||Issn 2250-3005(online|| ||February|| 2013 Page 13


i) MR-LQSR Protocol
Protocols for multi-hop routing in ad-hoc and WMNs proposed are DSDV, OLSR, TORA, and AODV DSR etc.
Protocols are mainly of two types i.e. on-demand/link state routing protocol better for mobile nodes and table-driven routing
protocols for static nodes. Link Quality Source Routing (LQSR) is a source routed link state routing protocol which supports
link quality metrics. LQSR implements all basic Dynamic Source Routing (DSR) protocol [5] functionality including route
request, route reply, route maintenance and feature of using one IP address as nodes home address for all communication
while in the ad-hoc network, in-turn each node independently assigns locally unique interface index to each of its network
interfaces. MR-LQSR is a combination of LQSR protocol with metric called WCETT. From [6] MR-LQSR has following four
components.
 Neighbor Discover
 Link Weight Assignment
 Link weight information propagation
 Path finding
The main difference with DSR and LQSR is DSR always tries to route through shortest path, whereas MR-LQSR uses
WCETT to assign the weights to links and does not follow shortest path. Shortest path routing performs worse many times.
Since MR-LQSR is used in layer 2.5 architecture which does not require to modify the protocol of layer above, uses 48-bit
virtual Ethernet address in its headers including source route, source reply and route error packets instead of 32-bit IP address.
Mesh Connectivity Layer (MCL) from [7]which operates as layer 2.5 is a driver which generates LQSR packet by inserting an
additional LQSR header into Ethernet packets received from the network layer as illustrated in fig.1+

Fig.1 Ethernet Packet from IP layer , LQSR
Packets generated by MCL

ii) Routing Metrics
Various routing metrics for operation in WMN is proposed and evolved from Hop Count (Hop) to Per-hop Round
Trip Time (RTT), Per-hop Packet Pair Delay (PktPair), Expected Transmission Count (ETX) and Weighted Cumulative
Expected Transmission Time (WCETT) with different intuitions in the way of improving network performance. We have
implemented ETX, PktPair and WCETT on our testbed and observed the throughput performance.

Expected Transmission Count (ETX)
ETX is the expected number of MAC layer transmission that is needed for successfully delivering a packet through a
wireless link. This metric is good in judging paths [8] by assigning large weights for long and lossy paths. ETX is derived as
packet loss rate in both the forward and reverse direction denoted by p fandpr respectively and then computes the expected
number of retransmission required to deliver the packet successfully to the destination as the IEEE 802.11 MAC retransmits a
packet whose transmission is not successful.. The equation 1 below illustrates the computation of ETX.
1
ETX  --------------1
(1  p f )(1  pr )
The equation also implies that ETX metric is bi-directional, that is the metric from node x to node y is same as metric from
node y to node x. The weight of a path is defined as the summation of the ETX’s of all links along the path.ETX does well in
homogeneous single-radio, but does not perform well in environments with different data rates.
Per-hop Packet Pair Delay (PktPair)
PktPair metric concept was thought off to correct the problem of distortion of RTT measurement due to queuing
delays. To calculate this metric, a node sends two probe packets back-to-back to each neighbor every second. First probe
packet is small and second one is large. The neighbor calculates the delay between receipt of first and the second packets and
reports this delay back to the sending node. The sender maintains an exponentially weighted moving average of these delays
for each if its neighbors. The objective of Pktpair routing algorithms is to minimize the sum of these delays, if due to high loss



rate second probe packet requires retransmission by 802.11 ARQ then delay measured by the neighbor increases. The
disadvantage of this metric is overheads is even greater than RTT and not completely immune to self-interference.

Expected Transmission Time (ETT)
From [4] ETT of a link k is defined as expected MAC layer duration required for a successful transmission of a
packet at link k. The weight of a path p is simply the summation of the ETTs of all the links on the path. This routing metric
improves the WMN performance by considering the differences in link transmission rates. The relationship between the ETT
of a link k and ETX can be expressed as shown in equation 2.
s
ETTk  ETX k ------------------------2
bk
Where bk is the transmission rate of link k and s is the packet size. The drawback of ETT is that it still does not
capture the inter-flow and intra-flow interference in the network completely. i.e. ETT may choose a path that only uses single
channel, even though a path available with more diversified channels with less interference or higher throughput.

Weighted Cumulative Expected Transmission Time (WCETT)
WCETT metric was proposed to reduce the intra-flow interference, means to reduce the number of nodes operating
on same channel on the path p, that is used to transmit the packets. Thus obtain high throughput in a multi-radio, multi-hop
WMN. This metric does not consider the effect of inter-flow interference. One of goal of MR-LQSR protocol design is path
weight should increase as more number of links are added to an existing path and that is fulfilled by this metric by
adding/increasing the ETT as the links are added. It is as well as end-to-end delay experienced by a packet travelling along the
path. Thus, for a path of n hops
n
WCETT  (1   )*  ETTi   * max X j ---3
1 j  k
i 1

In equation 3  is a tunable parameter subject to 0    1 and X j is the maximum number of times the channel j appears
along the path p, which captures the intra-flow interference, because WCETT metric assigns lower weights to the paths that
have more diversified channel assignments on their links.

We can interpret the equation 3 in two parts. First term is the sum all transmission times of all hops in the network,
which reflects the total resource consumption along this path. The second term reflects the set of hops that will have
mostimpactinthe throughput of this path. The weighted average can be viewed as an attempt to balance the two terms. WCETT
selects links based upon their loss rate and bandwidth, without regard to channel diversity if  =0 is set. It can also be
observed that WCETT is only metric that selects links based on both loss rate and bandwidth and considers channel diverse
paths unlike ETX.

3. Methodology
Since it is found suitable that the combination of LQSR protocol, WCETT metric for a multi-hop, multi-interface
WMN along with MCL which creates an ad-hoc routing framework is used in our experimental setup

Fig.2(a) Topology of the WMN



(b)
Fig.2(b) Schematic view of the testbed.

Fig. 2(a) shows a 10-node WMN testbed setup, located on the second floor of ED building with nodes placed in the offices,
conference rooms and labs. Unlike wireless-friendly cubicle environments, the building has rooms with floor-to-ceiling walls
and solid wooden doors. All the nodes were placed in fixed locations and never moved during testing. Node density was kept
fairly high to enable variety of multi-hop path. Our testbed spans 37 meters (120 feet approx) in length and 15.5 meters (50
feet) in width.Desktop PCs used in the testbed are 1.87 GHz Intel Core Duo processors with 2 GB memory with XP SP2 OS.
All the experiments were conducted on IPv4 using statically assigned addresses. Combination of single and two radio nodes is
used in our testbed, i.e. 4 nodes with two radios and 6 with single radio, with an intent to test the behavior of the network with
the combination of single and two radio nodes.Each node is equipped with combination of two 5 GHz and 2.4 GHz operating
band wireless adapters or with any one adopter such as D-Link DWL-520+ of IEEE 802.11b, Netgear WG-311 v3, D-Link
DWA-510 of IEEE 802.11g and Proxim Orinoco gold, ZyXEL AG-225H v2 of dual band that can operate on IEEE 802.11a&
IEEE 802.11g standards.Our testbed comprises of two networks (groups of nodes) i.e. nodes are grouped into N/W-A colored
with green and N/W-Bcolored with blue.Consider fig. 2(a) nodes with just circle either green or blue color is with single
interface and nodes with circle either green or blue with dots outside the circle is nodes with two interfaces.SSIDs for wireless
adapters belonging to nodes of N/W-A is configured as MyMesh for first adapter and MyMesh9 for second adaptor and for
nodes having single adapter, SSID configured is MyMesh. In N/W-B network all nodes are equipped with single wireless
adapter, configured with SSID MyMesh3, only gateway node of N/W-B is installed with two radios one configured as
MyMesh3, second adapter configured as MyMesh9. It can observed that adopter with SSID MyMesh9 is present in both the
gateway nodes of N/W-A and A/W-B which is responsible to bridge two networks here.
The static IP address for the nodes is assigned as given below
N/W-A - Starting from 10.10.11.9 to 10.10.11.16
N/W-B - Starting from 10.10.11.18 to 10.10.11.21
This IP address for a node is used to identify a node in the network, irrespective of the number of hardware adapters used as
part of the mesh network. Also following static address was assigned for the MAC of Virtual adapters of the nodes for better
monitoring while testing.

N/W-A - Starting from 74-99-99-99-AA-00 to 74-99-99-99-AA-06
N/W-B - Starting from 74-77-77-77-AA-00 to 74-77-77-77-AA-03

Fig.2 (b) illustrates the connectivity between the nodes of twonetworks. Since the nodes are combination of single
and two radios, nodes in the testbed is so arranged, such that any node of one network can communicate with any node of
other network.
Except configuring to ad hoc mode and fixing the frequency band & channel number, default configuration for the
radios is used. In particular, all the cards perform auto rate selection and have RTS/CTS disabled. It was ensured that there
were no other 802.11a/b/g users during the experiments.



4. Results
The software tools used in the testbed for measurement of throughputs between nodes with path lengths of one, two,
three and more hops were Iperf, Microsoft Network Monitor. Ping to check the connectivity and GNU-Plot to plot the obtained
results.

Single-Radio Experiment
Firstly the experiments with single radio on a best one hop path length with IEEE 802.11g and IEEE 802.11a was
conducted. Both the standards are of specification 54 Mbps data rate fig.3 shows a plot which illustrates the behavior of the
adopters. Procedure of the experiment is,a three minute tcp transfer was carried out between selected nodes first between a pair
of nodes with IEEE 802.11g adopters then with a pair of nodes with IEEE 802.11a adopters.

Fig.3 Comparison of throughputs between 802.11a linkand 802.11g link

It is observed that 802.11a link has offered a maximum throughput approx. 21 Mbps and transferred 314 MB of data whereas
802.11g link has given only 4.5 Mbps approx. as maximum throughput and had transferred only 44.8MB of data. Though both
the standards are of same data rate, 802.11g tends to be slower link compared to 802.11a link.

Impact of Band and Channel Assignment
An assumption while designing the MR-LQSR routing protocol is, if multiple radios used in a node then they should
be configured to on-interfering channels. Though it is done we can observe that performance is poor in the first set of
experiment. Table 1 show that data obtained from the experiment.

Receiver Sender-1 (Average Sender-2 (Average
(NR Lab) Throughput) Throughput)
Orinoco Adapter tuned to ch 64 Adapter tuned to ch 36
ZyXEL
Two 802.11a
Individually 21514 Kbps 12275 Kbps

Simultaneously 6206 Kbps 7970 Kbps
Drop in Throughput 71% 35%

Table1. Comparison of throughput between nodes with 802.11a interfaces
i) First Set of Experiment
Procedure of the experiment is three nodes all with IEEE 802.11a adopters are used. As shown in fig.2 (a) first node
is placed in conference room and second in NR lab and third in e-class room. All rooms are adjacent to each other. First
(Sender-1) and third (Sender -2) node is used always to send packets and second node which is in the middle at NR lab is used
always to receive the packets from the other two nodes. Both the sender receiver pairs of adopters in the nodes are configured
to non-interfering channels as shown in table 1.


Firstly tcp traffic was transferred for two minutes from first node only and received by NR lab and it was noted that
21514 Kbps was received. Secondly same experiment was done with third node with first node made idle and traffic received
at NR lab node and noted that 12275 Kbps of average throughput received. Finally from both first node and third node tcp
traffic was transferred for two minutes simultaneously and packets received at NR lab node. Now from table 1, it can be
observed that at the first sender the average throughput is only 6206 Kbps and from second sender it was 7970 Kbps wherein
the drop in throughput are 71% and 35% respectively compared to the throughputs with experiments when conducted
individually.

ii) Second set of experiment
Here sender-2, i.e. node 3 is now equipped with 802.11g adopter instead of 802.11a, also receiver adopter at node in
NR lab changed to 802.11g and configured to channel 10 as shown in Table 2
Again same set of previous experiments was carried out and results are noted as shown in table 2. It is observed that
drop in throughput is reduced significantly to 10% and 6% respectively.

Receiver Sender-1 (Average Sender-2 (Average
(NR Lab) Throughput) Throughput)
ZyXEL Adapter tuned to ch 36 Adapter tuned to Ch 10
D-Link
802.11a
Individually 802.11g 12345 Kbps 4698 Kbps

Simultaneously 10990 Kbps 4387 Kbps
Drop in Throughput 10% 6%

Table 2. Comparison of throughput between nodes with 802.11a and 802.11g a interfaces

Above two sets of experiments illustrates that though we use multiple interfaces with non-interfering channels, even then they
interfere and performance is very poor. This drawback is overcome in the second set of experiment by using multiple
interfaces but of different bands operating in 2.4 GHz and 5 GHz respectively in a node.

Experiment with Single/Two Radio with WCETT, ETX and PktPair Metrics
Test on WCETT, ETX and PktPair a shortest path routing was conducted to judge which metrics could be suitable for
wireless mult-mesh network.Procedure of the experiment is, firstly tcp traffic transfer for 180 seconds for a path length of 3
hops with all the four nodes equipped with single 802.11g adopter was conducted. Secondly all the nodes are equipped with 2
interfaces one with 802.11a and another 802.11g was used and same experiment was repeated. Results are tabulated in table 3.
From this experiment it can be observed from table 3 that WCETT metric even in single radio scenario does well
getting 20% and 65% more throughput compared to ETX and PktPair metrics respectively. This is because WCETT metric
considers not only loss rate but also link bandwidth while assigning weight to the links. Due to this WCETT sometimes may
select longer paths than ETX, even then results in better throughput. Nodes with two radios, no-doubt WCETT metric along
with LQSR meant for multi-radio scenario does better then single radio performance getting 140% and 540% throughput more
than ETX and PktPair metric in two radio scenario. Fig 4 shows the behavior of WCETT metric with ETX and PktPair both in
single radio and two radio scenario graphically.
Single Radio Percentage Two Radio Percentage
Experiment Increase Experiment Increase
Median Median
Throughput Throughput
WCETT 1240 Kbps 20% increase than ETX, 65% increase than Pkt Pair 2523 Kbps 140% increase than ETX,
540% increase than PktPair
ETX 992 Kbps 1031 Kbps
Pkt Pair 434 Kbps 394 Kbps

Table 3 Throughput with metrics WCETT, ETX and PktPair on Single/Two Radio nodes



Fig. 4 Comparison of median TCP throughput with WCETT, ETX and PktPair on Single/Two radio nodes

Fig. 5 & 6, illustrates the behavior of metrics ETX and WCETT with single radio and two radio node on path length
of one, two and three hops.

Fig. 5(a) ETX Single radio throughput comparison for single, two and three hop length

Fig. 5(b) ETX two radio throughput comparison for single, two and three hop length

From fig. 5(a) and 5(b) it can be observed that single hop throughput is almost the same, whereas the two hop path
performance has reduced and the three hop performance also has no significant increase in performance. This is because ETX
tries to route packets through 802.11g links which is longer range slower link even though higher throughput paths are
available. Hence ETX uses the second interface in sub-optimal manner.

Fig.6(a) WCETT single radio throughput comparison for single, two and three hop length



Fig. 6(b) WCETT two radio throughput comparison for single, two and three hop length

From fig 5(a) and fig. 6(a) when compared the throughputs are almost same. i.e. in single radio ETX does well but in multi-
radio scenario ETX performance is worse but WCETT metric, from fig. 6(b) the two hop path length performance is almost
double when compared to single radio performance. This is because WCETT uses always 802.11a link and uses 802.11g link
only when it is beneficial to it and WCETT uses the second interface as an additional resource given unlike ETX using it in
sub-optimal manner. Comparing fig. 6(a) and 6(b) three hop length performances has improved with two radio node compared
to single radio. Fig.7 shows performance improvement in percentage compared to single radio performance with WCETT
metric for various paths.

Fig. 7 Improvement in median throughput of two radio nodes over single radiofor various path lengths with WCETT
as the metric.

Single hop length performance improvement is not considered because of no channel diversity. Path length greater than 3 hop,
has only 30% improvement in performance, this is because on longer paths TCP performs poorer because of the following
reason.
 Increase in round trip time
 Higher probability of packet loss due to channel errors
 Contention between hops that are on the same channel.

Impact of Channel diversity (  )
WCETT metric is the weighted average of following two quantities
 Sum of ETTs of all hops along the path with a weight of (1-  )
 Sum of ETTs on the bottleneck channel, with a weight of 
Experiment with path length of two, three and more hops were conducted by transferring TCP traffic for 180 seconds with the
value of  = 0, 0.1, 0.5 and 0.9 set and readings are as noted in table 4 to measure the impact of channel diversity.

Table 4 Comparison of median TCP throughput for various pathlengths at various  values



Fig.8 shows  has an impact on throughput of connections of a specific path length. It is very clear that performance at
 =0.5 is better for all path lengths. At  =0 the performance is poor for all path lengths. The reason could be links used by
the metric at  =0 may be 802.11g link which yields lower throughput as seen in previous experiment and for paths longer
than 3 hops the channel diversity does not provide significant benefit.

Fig. 8 GraphComparison of median TCP throughput for various pathlengths at various  values

Parallel WMN
From previous experiments it is observed that in case of using multiple interfaces then if same band interfaces are
used then lot of interferences occur to extent that throughput drops to approximately 70%. This is because from [1] it is seen
that activating multiple radios in a node lead to degradation in performance due any of the following reason.
 Board Crosstalk
 Radiation leakage
 Inadequate distance separation between several antennas.
[1] Shows that by taking appropriate measure like shielding of wireless cards to reduce radiation leakage, custom made
platform to cancel crosstalk and adequate distance of 1 meter between the antennas, the performance can be improved to as
much as 100%. [2] has shown that throughput increases as we use more number of interfaces, but limits to 5. Using more than
five interfaces in a node leads to degradation of performance.

Impact of Throughput performance on parallel networks
In fig. 2(b), two mesh networks are set to operate in parallel i.e. N/W-Aand N/W-B. The testbed consists predominantly single
radio nodes, with a few two radio nodes used. To test transfer flow behavior between two networks, three minute TCP traffic
transfer was conducted with path length of single, two, three hop lengths and also same experiment on a single WMN for
comparison was done. Results are as tabulated in Table 5.

Fig.9 shows the behavior of throughput performance between nodes in a single network and across two networks. It
is observed that both throughputs are almost same and closely follows except a small difference

Path Length Median Media Throughput Between Difference in Throughput
Throughput nodes in Two Network
Between nodes in
Single Network
Two-hop 2032 Kbps 2228 Kbps 196 Kbps
Three-hop 583 Kbps 712 Kbps 129 Kbps

Table 5 Comparison of throughputs between nodes in a single network nodes and nodes across two networks



Fig. 9 Comparison of throughput between nodes in single network and nodes across network for a path length of
single, two and three hops

5 Conclusions
Routing protocol that can operate with multi-radio concept like LQSR and efficient metric that utilizes the multi-
radio resources beneficially like WCETT routing, also routing metric should consider both inter-flow and intra-flow
interferences while routing the packets. Here WCETT does not consider inter-flow interference and hence can transmit packets
to the congested route also, otherwise WCETT is the best suited metric for atmosphere like multi-hop, multi-interface and
multi-mesh wireless network.As per [2] if maximum of 4 physical network interfaces can be used in a node with the measures
taken to shield the network interfaces and by using custom made box to avoid crosstalk and adequate separation between
antennas, thenwith multiple radios of different bands in a nodecan perform best without interference and is possible to set up
multiple, say 4WMNs that can operate in parallel, such that we can extends the coverage area of Mesh network, thus
enhancing the capacity of WMN.

References
[1] Joshua Robinson, Konstantina Papagiannaki, Christophe Diot, Xingang Guo and LakshmanKrishnamurthy. “Experimenting with a
Multi-Radio Mesh Networking Testbed,” in 1st workshop on Wireless Network Measurements WinMee, Trento Italy, 2005.
[2] Chi Moon Oh, Hwa jong Kim, Goo Yeon Lee and Choong Kyo Jeong , “A Study on
the Optimal Number of interface in Wireless Mesh Network” International Journal of Future Generation Communication and
Networking, IJFGCN Vol. 1, No. 1 pp 59-66, Dec 2008.
[3] Yn Zhang, Jijun Kuo and Honglin Hu. WIRELESS MESH NETWORKING, Boca Raton, New York: Auerbach Publications 2007.
[4] Yaling Yang, Jun Wang and Robin Kravets “Designing Routing Metrics for Mesh Networks,”IEEE Workshop on Wireless Mesh
Networks (WiMesh), Santa Clara, CA, 2005.
[5] David B. Johnson, David A. Maltz, and Josh Broch, “DSR: The Dynamic Source Routing Protocol for Multi-hop Wireless Ad-hoc
Networking” in Ad Hoc Networking, 1/e. Charles E. Perkins, Addison-Wesley, 2001, pp. 139-172.
[6] Richard Draves, Jitendra Padye and Brian Zill. “Routing in Multi-Radio, Multi-Hop Wireless Mesh Networks,” Mobicom, 2004, pp
114-128.
[7] Richard Draves, Jitendra Padye and Brian Zill. “Comparison of Routing Metrics for Static Multi-Hop Wireless Networks,”
SIGCOMM’04, 2004, pp 133-144.
[8] Douglas S J De Couto, Daniel Aguayo, Jhon Bicket and Robert Morris, “A High-
throughput Path Metric for Multi-hop Wireless Routing.” Journal of Wireless Networks,
Vol. 11, Issue 4, pp 419-434, Jul 2005.


International Journal of Computational Engineering Research(IJCER)

More Related Content

What's hot (20)

Similar to International Journal of Computational Engineering Research(IJCER) (20)

International Journal of Computational Engineering Research(IJCER)