VOL. 2, NO.

10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

Investigation and Addressing Unfairness in Wireless Mesh Networks

1 2, 3

M. I. Gumel, 2N Faruk , 3A.A Ayeni Huawei Technologies, Kano, Nigeria

Department of Telecommunication Science, University of Ilorin, Nigeria

Corresponding Author: faruk.n@unilorin.edu.ng

ABSTRACT
Unfairness is one of the major problems in WMN as each node operates not only as a host but also as a router, forwarding packets on behalf of other nodes that may not be within the direct wireless transmission range of their destinations and thereby adding delay and affecting the throughputs of the node. Addressing end-to-end unfairness in wireless mesh network is necessary for the network to be scalable otherwise; the expected network scalability will not be met. The unfairness investigation shows that the end-to-end unfairness in a wireless mesh network will affect the network scalability. Modification to the network layer has been proposed to include an additional queue, a weighted fair scheduler and a classifier to separate locally generated flow from the relayed flows at each node. This will make the traffic flows from all the sources in the network to achieve equal throughput at the destination irrespective of the number of hops traversed. Another unfairness issues is also investigated at the MAC layer as a result of adopting existing single hop MAC protocol to be used in this multi-hop mesh network.
Keywords: Wireless Mesh Networks (WMN), Medium access Control, MAC layer, Qos

1. INTRODUCTION
Wireless mesh networking (WMN) is another option of scaling wireless network. A wireless mesh network is a type of wireless networks in which each node can communicate directly with one or more peer nodes, is a form of ad hoc network that form mesh of wirelessly interconnected nodes. The difference between this type of network and the conventional ad hoc network is the nature of packets movement in the network which is always from the clients to the gateway node or vice versa in wireless mesh networks, while the packets movement in ad hoc networks is between arbitrary pair of the network nodes [1]. In wireless mesh networks the host nodes (typically stationary) can also serve as routers to forward the clients traffic in a multi-hop fashion to the destination (Normally gateway nodes connected to the internet) when the network is deployed to provide internet access. Wireless mesh networks offer advantages over other wireless networks; these include easy deployment, greater reliability, self-configuration, self-healing, and scalability. The nodes in the network are able to establish and maintain mesh connectivity automatically [2].Wireless mesh networks technology has drawn considerable attention as a promising broadband access technology despite the increase in the number of internet access technologies in fig 1 below. Very large areas can have access to broadband wireless using wireless mesh networks without need of costly infrastructures. Wireless mesh networks can also be employed for wide variety of applications such as cellular radio access networks or wireless local area networks (WLAN) hotspot multihopping, citywide surveillance systems, wireless sensor networks (WSNs), broadband home and office Indoor networking, intelligent transport system networks, community and neighbour networking, micro base station backhaul and many others [3,4].

Wireless mesh network is a multi-hop network in which each node operates not only as a host but also as a router, forwarding packets on behalf of other nodes that may not be within the direct wireless transmission range of their destinations [2]. Each mesh node in the network will have to be able to forward more than one traffic flow, the traffic flows generated by the clients attached to it ( locally generated traffic flow) and the traffic flows of the other nodes (relayed traffic flows). The network resources (buffer space, bandwidth) are always finite in the any network to provide broadband access, so resources may have to be allocated and scheduling decisions have to be made [1] and there is possibility of delay in any system that involves waiting, so traffic flows that pass through many hops to their destination will experience more delay, and the end-to-end delay varies with the number of hops to the destination. The end-to-end unfairness among the traffic flows in a wireless mesh network results from the queuing system in the network, in which packets are buffered while waiting to be transmitted [5]. Traffic flows achieve different throughput at the destination because of the different number of nodes they pass to the destination, the delay sum up along the path and is higher in some traffic flows than the others. Another unfairness issue evolve when more than one node carrying different number of traffic flows tries to send traffic to the same destination, the conventional IEEE MAC layer protocol treats the nodes equally irrespective of the number of flows passing through each node. This results in unfairness between the traffic flows in the network. The throughput capacity of wireless mesh networks and the network performance will be increased if

514

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

the issue of unfairness behaviour of the flows in the network is addressed. The fair sharing of network resources can be achieved by giving the flows from distance source nodes crossing many network hops to the destination priority at each node queue over the flows from hops close to the destination (gateway). This reduces the end-to-end delay of the traffic flow and makes all the source nodes in the networks to have equal share of the network resources irrespective of their distance from the destination.

in the networks. But separating the each traffic flow means providing a separate queue at both network and MAC layer for each flow; this will make the node to be complicated and increase processing delay at the node, which will reduce the network efficiency. Another problem is the number of traffic flow through each node in the network which is not predictable. While working on this research, we came across [6] addressing unfairness issue in multi-hop wireless mesh. They have limited their research to LAN using no routing protocol. They have proposed a dual queue algorism to be implemented above the MAC layer to separate locally generated traffic flow from relayed traffic flow to provide fairness, which is related to our approach of addressing unfairness but different methodology and implementation. The end-to-end unfairness among the traffic flows in a wireless mesh network results from the queuing system in the network, in which packets are buffered while waiting to be transmitted [5], and the network layer responsible for end-end-end transmission of traffic flows [6, 7] uses a queue where in the case of a wireless mesh network both the locally generated traffic flow and the relayed traffic flow are buffered. So to address the end-to-end unfairness modification of the existing network layer to provide dual queues to separate relayed traffic flow from locally generated traffic flow will be a simplest and efficient solution as shown in this research. 2.1 CONVENTIONAL QUEUING SYSTEM AT THE NETWORK LAYER The queuing system at the network layer is shown in fig.2 below. The system uses only one queue with a buffer to hold the packets while waiting to be forwarded. The scheduling in the system is based on first in first out (FIFO) with tail drop [6]. When the rate of arrival of the packets is greater than the rate of the transmission, the buffer will be filled and starts overflowing. The tail drop used in the system means the packet that arrives last will be the one to be dropped [7].

Fig.1: Wireless Mesh Network for broadband Internet access [1]

The aim of this research is to investigates (using OPNET Modeller 14.5) the end-to-end throughput unfairness behaviour of traffic flows in multi-hop wireless mesh networks based on the position (the number of hops passed to the gateway) of the traffic source in the network relative to the gateway; the investigation is performed at the network layer which is responsible for end-to-end packets transmission (from source to destination across multiple hops).

2. RELATED WORK
Investigation and addressing unfairness in wireless mesh networks is an important issue currently under research in the field. [5] In their research on fairness and multi-hop wireless networks have studied the possibilities of implementing various queuing schemes in multi-hop network to address the issue of unfairness and QOS provision in the network. They have investigated how the various queuing scheme will suit the network layer and MAC layer in the network by studying the effect of separating traffic flows at the network layer and the effect of separating the flows at the MAC layer. After studying the conventional method of using single queue at the network layer and single queue at the MAC layer, they have studied several solutions to unfairness problem, either by using dual queues (to separate locally generated traffic flow from relayed traffic flow) at the network layer and single queue at the MAC layer, using multiple queues (to separate each traffic flow) at the network layer and single queue at the MAC layer, and using multiple queues at both network and MAC layers. They have finally concluded that, using multiple queues to separate each traffic flow at both network and MAC layer will provide better fairness

Fig.2: Conventional queuing system at the network layer

515

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

Employing this system to work in wireless mesh networks means both the locally generated traffic flows (LTF) and relayed traffic flows (RTF) will share a single queue and be treated as first-in first-out. 2.2 SIMULATION ENVIRONMENT SET UP The simulation environment has been set up using an empty OPNET initial topology and campus network scale, with 10 x 10 km2 area. The mobile adhoc network (MANET) technology is used throughout in this research. Wireless mesh network scenarios have been deployed in the OPNET Modeller environment and simulations were performed. 2.3 SIMULATION PARAMETERS All the simulations have been done using the MANET raw traffic generation, and each node that is set to transmit is assumed to have unlimited amount of traffic to send to the gateway node. The offered load by each node in the network is increase by changing the packet inter arrival rate in the node. For an offered load of 100 packets per second, the packet inter arrival time is set to 0.01; this means that the node generates 100 packets per second ). The start packet generation time was set to 5s ( while the end time set to the end of the simulation. Each simulation was run for 5mins. Table1.0 below shows the other parameters used. TABLE 1.0: Simulation Parameters PARAMETERS VALUES Data rate Packet payload Transmission Range RTS Threshold Transmit power Tx Receiver Sensitivity Rx Fragmentation Buffer size Simulation time Modulation format 11Mbps 12000bit 1250m 250 byte 7dBm -95dBm none 256000bits 300sec DSSS

flows from individual mesh node, four nodes have been placed at the gateway position to receive traffic flows from individual mesh node in the network. Flows from node_1, node_2, node_3 and node_4 are to be received by node_0, node_5, node_6 and node_7 respectively placed at the same position (gateway position).

Fig.3 Linear WMNs scenario with queuing system in the network layer of the mesh nodes

In wireless mesh networks, the mesh nodes act as wireless clients and mesh routers as well. A mesh node can send its generated traffic (the traffic from the clients connected to it, not shown in the figure) as well as other mesh nodes traffic. So node_1 forwards traffic of node_2, node_3 and node_4 in addition to its own traffic. Node_2 forwards its own traffic and also traffic of node_3 and node_4. Node_3 forwards traffic of node_4 and its own traffic, as shown in fig.3 above. If the mesh nodes are assumed to be sending equal number of packets per second. Node_3 would be sending two times packets per second of node_4; node_2 would be sending three times packets per second of node_4, while node_1 would be sending four times packets per second of node_4. The mesh network in fig.3 above has been simulated with each mesh node set to transmit in the network. The offered load (generated traffic) by each mesh node started from 10 packets per second (0.1 packets inter arrival rate) then increased in step of 10 to offered load of 100 packets per second (0.01 packets inter arrival rate). 2.5 SIMULATION RESULTS To investigate the end-to-end unfairness in the network the simulation results are taken from the network layer (in IP traffic) at the destination. The average throughput of the individual node traffic flows for each simulation (using different offered load) were plotted against the offered load (traffic generated), as shown in figure 4 below:

2.4 NETWORK SCENARIO SIMULATION The simulation environment has been setup using the simulation parameters shown above. Four nodes (node_1, node_2, node_3 and node_4) as shown in fig.3 below have been placed in such a way each node only transmits and receives from the node next to it. This is to ensure the packet transmitted to the destination traverse the network using multi-hop. The gateway node is the destination of all the transmitted traffic. To differentiate

516

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

Fig.4: Average per flow throughput (packet per second) with increasing offered load

Fig.5: Average (in IP Traffic Received (packet per second)) using offered load of 10 packets per second.

Fig.4 given above shows the average throughput of the individual nodes with increasing offered load. Each node achieves 100% throughput of its generated traffic when the total offered load in the network is low (all the generated traffic from the nodes in the network get through). When the offered load in the network increases, processing delay also increases at the queuing system, this makes throughput of some nodes to drops. This happens after offered load of about 40 packets per second as shown in fig.4. At each mesh node the locally generated flows tend to dominate the queue and the relayed flows from other nodes that join the queue later experience high delay and more chance of being dropped. This makes the mesh nodes to have different throughput based on their relative position to the gateway since their end-to-end delay sum up and affect throughput at the gateway node. Fig.5 given below shows the simulation result using offered load of 10 packets per second where each node achieves 100% throughput of its generated traffic. The result is the average throughput of each flow over the simulation time. The initial low throughput up to about 50s is to do the simulation settling to its equilibrium value and the result is average over the simulation time.

Fig.5 given above shows that the nodes average throughput reaches up to about 15 packets per second while the each node generates only 10 packets per second; the remaining packets are due to routing traffic in the network as shown in fig.6 below.

Fig.6: Average (in AODV Routing Traffic Received (packets/second))

The average value of IP traffic received is higher than the expected 10 packets per second generated by each mesh node, this is because of the routing overhead caused by the routing traffic exchanges (e.g. RREQ packets, RREP packets and RRER packet in case of route error) [9] when using AODV routing protocol. The average AODV routing traffic received at the network layer of each individual node at the destination. To compare transmitted and received traffic at the same level to show that they are the same value, the average in IP traffic sent is also plotted using the same offered load as shown in fig.7 below where node_4 at the end of the linear network transmits about 15 packets per second, the same with number of packet received at the

517

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

destination shown in fig.5. Node_3 transmits twice of node_4 transmission, node_2 three times while node_1 four times.

issues at the network layer of the mesh nodes in wireless mesh networks can be easily observed. The network scenario used also confirmed that the throughput of each mesh node depends largely on the position of the mesh node relative to the gateway (destination). The position corresponds to the number of hops that packets from the mesh node pass before reaching the gateway. Unless the unfairness issue is addressed in the wireless mesh networks, the throughput of the mesh node in the network will depend on the position of the node relative to the gateway. So wireless mesh networks will only be scalable if the unfairness issue is being addressed.

3. ADDRESSING THE END-TO-END UNFAIRNESS PROBLEM IN WMNs

Employing the traditional queuing policy to work in wireless mesh networks results in unfairness among the traffic flows from individual mesh nodes. This is because at each mesh node the relayed traffic flows and the traffic flow locally generated share the same queue. The unfairness is well pronounced when the traffic in the network is high. The end-to-end unfairness issue in wireless mesh networks can be addressed since the root of the problem is identified. A scheme is needed to give traffic flows from distance nodes priority over the locally generated traffic at each transit node. The scheme should reduce the flows end-to-end accumulated delay and increase the traffic flow throughput at the destination. Network layer modification is proposed in this research to include an additional queue to separate locally generated traffic flow from the relayed traffic flows instead of using only one queue. The modification will include the use of a weighted fair scheduler and a classifier. The classifier will classify the incoming packet and forward it to the appropriate queue and the weighted fair scheduler will use the scheduling formula in equation (1) and (2) below to forward the packets from the two queues in a weighted fair fashion. This system will reduce the relayed traffic flows end-to-end delay and increase the flows throughput at the destination. Fig.9 below shows the modified network layer queuing proposed with a classifier and a weighted fair scheduler, to be implemented in each mesh node.

Fig.7: Average (in IP traffic sent (packet per second)) using offered load of 10 packets per second

When each mesh node set to generate 100 packets per seconds the result obtained is shown in fig.8 below.

Fig.8: Average (in IP Traffic Received (packet per second)) using offered load of 100 packets per second.

For the offered load of 100 packets per second from each node, node_1 which is just one hop to the gateway (destination node) achieves 100% throughput of the traffic sent, measured at node_0 destination. Node_2 which is two hops to the destination achieves throughput of about 50% of the traffic sent measured at node_5 destination. Node_3 which is three hops to the destination achieves throughput of about 10% of the traffic sent measured at node_6 destination. Node_4 which is four hops to the destination achieves throughput of less than 10% of the traffic sent measured at node_7 destination. With the simulation results above, the unfairness

518

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

priority over locally generated traffic flow (to reduce its end-to-end delay) something must be done to prevent the locally generated flow from being starved when there is continues flow of relay traffic through the higher priority queue. This problem can be addressed by employing a weighted fair scheduling scheme, where the scheduler will schedule the packets from the queues for transmission in a weighted fair fashion to avoid starving the locally generated flows. In weighted fair queuing multiple queues are serve so as to share the link capacity in defined proportion amongst the queues [6]. The formula is needed since the scheduling rate of the queues is not constant. The scheduling rate of each queue depends on the total number traffic flows passing through the queues (locally generated and relayed traffic flow) [6]. The scheduling formula below, defines the scheduling proportion that will be giving to each queue by the scheduler. (1) LTQSR (%) = (1/ FT)*100 RTQSR (%) =100-LTQSR (2) Where, FT is the total number of flows, LTQSR and RTQSR are the Local traffic scheduling rate and relay traffic scheduling rate respectively. The formula above will be used by the scheduler to schedule packets from the two queues for transmission, and the relayed traffic queue will be given more percentage based on the number of flows in the queue. 3.3 ANALYSIS AND SIMULATION OF THE NETWORK SCENARIO USING THE PROPOSED MODEL The wireless mesh network scenario in fig.3 is also used to simulate the proposed model. The mesh routers are modified to use two queues, a classifier and a weighted fair scheduler at the network layer. The queuing at the network layer is shown in fig.11a below. When all the mesh nodes generate traffic to the gateway node, there is no relay traffic through node_4, so its relay traffic buffer is empty.

Fig.9.0: Proposed modification to Network layer in wireless mesh network

3.1 THE CLASSIFIER Differentiated services technology on the internet uses a classification process in which a classifier is used to classifier and aggregates all packets that need similar QoS into same class, per hop behaviour (PHB) [6]. This kind of classifier can be used in this work to classify the arriving traffic packets into either relayed traffic flow or locally generated traffic flow and forward to the appropriate queue. Upon receiving a packet from a flow, the classification process is performed, where classifier will examine some fields (hop count, source and destination addresses) in the packet routing header to identify the packet as either locally generated or relayed traffic flow. The process is shown in the flowchart in fig.10 below. The flowchart described the classification process at each intermediate node.

SourceNode

Start

SendTrafficFlow

Intermediate Node Yes

No

Istheflowlocally generated?

Relayedtraffic Queue

LocallyGenerated TrafficQueue

Fig.10: Classification Decision at the Intermediate Node

Fig.11a: Linear WMNs scenario with modified queuing system in mesh nodes network layer

3.2 THE WEIGHTED FAIR SCHEDULER Although the proposed scheme to address the end-to-end unfairness issue will give relay traffic flow

519

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

110 100 90

Local traffic Relay traffic

S c h e d u lin g R a te (% )

80 70 60 50 40 30 20 10 0 0 1 2 3 4 5 6 7 8 9 10 11

Four created weighted fair queue node models were renamed as node_1, node_2, node_3 and node_4, and the attribute of each node was edited to have two queues with different weight based on the node position (position 1-4) Node_4 queues set to have weight of 100 and 0, node_3 queues have 50 and 50 at both the queues, node_2 queues have 33.33 and 66.67 and finally node_1 queue have 25 and 75 weight for local traffic queue and relay traffic queue respectively. Each weighted fair queue node model was configured to have one traffic source attached to the local traffic queue, while the traffic flow from the node next to it will used the other queue for the relay traffic.

Total Number of Flows

Fig 11b: Scheduling rate of the queue as a function of number of flows

The simulation was performed with offered load of 100 packets per second (generated packets at each source node) at each mesh node and the buffer size in each node increased to avoid dropping of any packet. 3.4 SIMULATION RESULTS Fig.12 below show the simulation result of the proposed method of modifying network layer to use dual queue to provide fairness among the traffic flows in wireless mesh networks irrespective of the nodes position. The result shows that all the flows achieve equal throughput flow at the gateway node, instead of the nodes closer to the gateway node to dominate the network and starve the other nodes far away from the gateway when using the conventional single queue in each mesh router. The proposed method of using dual queue gives all the nodes equal opportunity of using the network resources, but the maximum throughput of each node decreases by where n is the number of hops pass along the path to the destination.

In fig 11a above, the total number of traffic flows through node_4 is 1, and then the local traffic queue scheduling rate will be 100% as shown in fig 11b. At node_3, the total number of the traffic flows through the node is 2, the nodes locally generated traffic and node_4s traffic flow passing through the node to the gateway node (relayed traffic flow). The relay traffic queue scheduling rate will be 50% and the local traffic queue scheduling rate will also be 50%. Always 50% of the nodes transmission will be from local traffic queue and remaining 50% is from relay traffic queue if node_4 is transmitting. Node_3 delay the transmission of its locally generated traffic flow to provide fairness. At node_2, the total number of the traffic flows including the relay traffic of node_4 and node_3 is 3. The relay traffic queue scheduling rate will be 66.67% and the local traffic queue scheduling rate will be 33.33%. Similarly, at node_1 the total number of the traffic flows including the relay traffic of node_4, node_3 and node_2 is 4. The relay traffic queue scheduling rate will be 75% and the local traffic queue scheduling rate will be 25%. This is the operation principle of the proposed method of enforcing fairness among the wireless mesh networks traffic flows from the mesh nodes at different positions. The relay traffic flow is given priority over locally generated flow at each transit node, this reduce its end-to-end delay and increase its throughput at the destination. The locally generated traffic flow is delay to avoid dominating the nodes transmission and weighted fair scheduler prevents it from being starved. In [10], the authors have worked on development of laboratory exercises based on the OPNET Modeller, they demonstrated the behaviour of different queuing disciplines used in differentiated services architecture. They have showed a step-by-step method of creating weighted fair queue using the OPNET Modeller predefined process and node model files. The proposed dual queue with weighted fair scheduler is model in each mesh router in similar approach with the work of [10].

Fig.12: Average throughput of individual node traffic using proposed double queue

Fig.12 above shows the traffic flows from the individual mesh node in the network, all the mesh nodes

520

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

achieve almost equal throughput. Each node generates 100 packets per second; the maximum achievable throughput result of each node is 25 packets per second using explained above. The result shows the average throughput of the individual node traffic flow over the simulation time, and shows each node achieves throughput higher than the expected 25 packets per second because of the overhead due to the routing traffic exchanges which increased the number of packets received at the destination as explained above. The simulation result shows that the maximum achievable throughput by each node in the multi-hop wireless mesh networks depends on the number of nodes that are participating in multi-hop forwarding (number of hops traversed to the gateway node) in the network, and is the same for all the nodes. Addressing end-to-end unfairness in wireless mesh network is necessary for the network to be scalable otherwise; the expected network scalability will not be met since the throughput of the node far away from the gateway is almost zero using the conventional method. The end-to-end unfairness can be addressed by modifying the network layer to include an additional queue to separate the locally generated traffic flow from the relayed traffic flow; this prevents locally generated traffic from dominating the channel and starving the relayed traffic. The main limitation of the proposed system is the throughput of the nodes closer to the gateway decrease compared to using conventional method of using a single queue at the network layer, but this is done to give other nodes far away from the gateway chance to send their traffic.

Fig.13: A WMNs scenario to investigate MAC layer unfairness

Node_4 and node_1 within the gateway node transmission range can directly send traffic to the gateway node. Node_1 sends its traffic flow and relay traffic of node_2 and node_3. Node_4 with no relay traffic; sends only its traffic. Node_4 with only one traffic flow (locally generated traffic) needs to send generated traffic of only 100 packets per second, while node_1 with three traffic flows needs to send traffic of up to 300 packets per second (including the relayed traffic of node_2 and node_3). Both node_1 and node_4 are within the gateway nodes transmission range, and they contend for the same medium to transmit. Using the conventional IEEE 802.11 DCF protocol both the node_1 and node_4 given equal opportunity of accessing the medium [8] irrespective of the number of traffic flows passing through each node. 4.1 SIMULATION SCENARIO OF THE WMN NETWORK

The wireless mesh networks scenario in fig.13 has been simulated with each node set to generate 100 packets per second. The simulation results are shown in the fig.14 and fig.15 below. The results have been taken at the MAC layer of node_1 and node_4. 4.2 SIMULATION RESULTS

4. INVESTIGATION OF MAC LAYER UNFAIRNESS IN WMNs

This section investigates another unfairness issue due to the nature of MAC protocol adopted to be used in wireless mesh networks. Wireless mesh networks adopt existing IEEE 802.11 DCF MAC protocol to be used as a medium access protocol [13]. The unfairness issue is due to the fact that the IEEE 802.11 DCF MAC protocol is designed to be use in single hop network [11] and the nodes in wireless mesh network communicate in multihop fashion. A wireless mesh network scenario in fig.13 below is used to investigate the MAC layer unfairness. All the mesh nodes (node_1, node_2, node_3 and node_4) in the network transmit to the destination (gateway) and each node is set to generate 100 packets per second. So the offered load G is 100 packets per second. The mesh nodes are arranged. For node_3 and node_2 to send traffic to the gateway, the communication must be through node_1; this is an example of multihop communication.

Fig.14: Average data traffic sent (packets/sec) by node_1 and node_4

The result in fig.14 above shows the average data traffic sent by node_1 and node_4. It shows that node_4 sends average of 100 packets per second (it achieves 100%

521

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

throughput of its generated packets), while node_1 is able to send maximum of about 210 packets per second which is less than the intended transmission of 300 packets per second. This is because all the 2 nodes are given equal opportunity to access the transmission medium and node_1 has to spare some time to receive traffic of node_2 and node_3, since it cannot send and receive at the same time. So node_4 sends 100 packets every second but node_1 can only send maximum of about 210 packets within that second. This 210 packets per second traffic is the traffic of three nodes (node_1, node_2 and node_3), while node_4 in the network send traffic of 100 packets per second alone, there is unfairness among the traffic flows in the network.

same medium based on the amount of traffic each node needs to send. The contention window allocated to a node is an interval with minimum (CWmin) and maximum value (CWmax). A node can be given priority over another node by allocating it small value of CWmin while the other node(s) are allocated large value of CWmin value. The node with large CWmin value will have a large backoff interval [20, 22]. Allocating different CWmin value to nodes based on the amount of traffic on each node relative to other node(s) will adjust the rate at which each node will be transmitting. By considering the network scenario in fig.13, node_1 and node_4 are contending for the same channel to transmit to the gateway node. The modified MAC layer protocol will give node_1 priority of accessing the channel over node_4 by allocating node_4 a large value of CWmin than node_1. Node_1 has 3 traffic flows, so the total number of flows to the gateway is 4. If node_1 is allocated a value of X for CWmin, then node_4 will be allocated a value of 4X for CWmin. This makes node_1 to have three times more access to the medium than node_4. 4.4 SIMULATION THE SCENARIO USING THE PROPOSED MODEL To test the performance of the proposed model the network scenario in fig.13 has been simulated with different CWmin values at node_1 and node_4 contending for the same channel. The default CWmin of 32 was set to node_1, while 4 times 32= 128 was set at node_4, to give node_1 priority of accessing the medium over node_4 (node_1 has small back off time). Fig.16 and fig.17 below show the simulation results using the proposed model. 4.5 SIMULATION RESULTS

Fig.15: Average delay (sec) by node_1 and node_4

Fig.15 above shows the average in wireless MAC layer delay of the 2 nodes, node_1 with large amount of traffic to send experiences higher delay of about 0.004sec, while the average delay experiences by node_4 is only about 0.001sec . The delay of node_1 increases because of the time spends by node_1 to receive traffic from node_2 and node_3. 4.3 ADDRESSING THE UNFAIRNESS ISSUE IN WIRELESS MESH NETWORKS MAC LAYER The MAC layer unfairness in multihop wireless mesh networks is as a result of giving the nodes trying to access a transmission channel equal opportunity irrespective of the amount of traffic each mesh node is sending. To address the issue of unfairness in the MAC layer, this research proposes a simple method, in which each mesh node will be given opportunity of accessing the medium based on the amount of traffic the node is sending relative to other nodes. The node with the largest amount of traffic will be given higher priority of accessing the channel over the other nodes. This can be done by modifying the 802.11 DCF MAC layer protocol to allocate different contention windows to the nodes accessing the

Fig.16: Average data traffic sent (packets/sec) using the proposed method

Fig.16 above shows the simulation result using the proposed method; the result shows an increase in traffic sent by node_1 with 3 flows compared with the

522

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

result obtained in fig.14 above using the conventional method. Using the conventional method node_1 sends maximum of 210 packets per second, while with the proposed method the node can send almost 245 packets per second. Allocating a large value of CWmin to node_4 with only 1 traffic flow increases the delay of the node in sending traffic to the gateway as shown in fig.17 below. This can have effect on the data traffic sends by the node since increasing the delay reduces the data traffic send, but still the node can send about 90 packets per second.

traffic flow traverse to the gateway node. The method improves the wireless mesh networks scalability.

REFERENCES
[1] J. Jun and M. Sichitiu , The nominal capacity of wireless mesh networks, 2003 Wireless Communications, IEEE Vol:10 Issue: 5 pp): 8 14 [2] I. Akyildiz, X. Wang, and W. Wang, Wireless mesh networks: A survey, in Computer Networks2005, vol. 47, no. 47, pp. 445-487. Retrieved on 31 Feb, 2010 [3] Y. Zhang and J. Luo, H. Hu Wireless mesh networking Architecture, Protocols and Standards, 2007 Wireless Networks and Mobile Communications Series Auerbach Publications, New York [4] Y, Zhang and J. Zheng, Security in Wireless Mesh Networks 2009 by Taylor & Francis Group, LLC Auerbach is an imprint of Taylor & Francis Group, an Informa business [5] J. Jun and M. Sichitiu. Fairness and QoS in Multihop Wireless Networks Department of Electrical and Computer Engineering North Carolina State University Raleigh, NC 27695-7911 [6] G. Tagg Differentiated Services on the Internet Unpublished Lecture notes Mar., 2010 Oxford Brookes, University, UK [7] D. Tung et al. Wireless Broadband Networks. 2009 Published by John Wiley & Sons, Inc. Hoboken, New Jersey [8] B. Forouzan Data Communications and Networking 2007 McGraw-Hil [9] C. Perkins Ad hoc On-Demand Distance Vector (AODV) Routing Network Working Group Request for Comments: 3561 Feb. 2003 Retrieved on 29 Feb, 2010 [10] T. Svensson and A. Popescu Development of laboratory exercises based on OPNET Modeler 2003 Thesis Master of Science Degree in Electrical Engineering with emphasis on Telecommunications and Signal Processing Blekinge Institute of Technology [11] A. Mukhherjee, S. Bandyopadhya and D. Saha Location Management and routing in Mobile Wireless Networks Mobile communication Series, Artech House Boston, London [12] S. Khurana, et al Effect of Hidden Terminals on the Performance of IEEE 802.11 MAC Protocol Wireless Communications, IEEE Vol:15 Issue: 5 pp): 8 24 [13] S. Szott, et al Impact of Contention Window Cheating on Single-hop IEEE 802.11e MANETs 2001 Colorado State University Ft. Collins, Colorado, U.S.A [14] E. Hossain, K. Leung Wireless mesh networking Architecture and Protocols 2008 Springer Science + Business Media, LLC 233 Spring Street, New York, NY 10013, USA [15] F. Li Yu and X. Li Gateway Placement for Throughput Optimization in Wireless Mesh Networks Department of Computer Science University of North Carolina at Charlotte, USA [16] J. Song et al Efficient On-Demand Routing for Mobile Ad-Hoc Wireless Access Networks 2004

Fig.17: Average delay (sec) using the proposed method

Fig.17 above shows the delay experienced by the mesh nodes using the proposed method. The result shows that the average delay for the node_1 decreases to almost 0.002secs with use of the proposed model compared to 0.004secs using the conventional DCF MAC protocol as shown in fig. 15 above. Adopting the IEEE 802.11 DCF MAC protocol to be used in wireless mesh networks results in unfairness among the traffic flows in the networks as the protocol is designed to be used in single hop network and give nodes equal opportunity in accessing the transmission channel [2]. Giving a node priority of accessing a channel over the other nodes based on the amount of its traffic load relative to other to traffic load of the nodes provides fairness among the traffic flows in the network. This can be done by allocating different minimum contention windows to each node based on its traffic load relative to the total traffic of the nodes accessing the channel.

5. CONCLUSION
The multihop communication in a wireless mesh network introduces unfairness among the traffic flows in the network and affects the network scalability. The proposed method of modifying the network layer to include an additional queue, a weighted fair scheduler and a classifier to separate locally generated flow from the relayed flows at each node makes the traffic flows from all the nodes in the network achieve equal throughput at the destination, this is irrespective of the number of hops each

523

VOL. 2, NO. 10, October 2011

ISSN 2079-8407

Journal of Emerging Trends in Computing and Information Sciences

2009-2011 CIS Journal. All rights reserved.

http://www.cisjournal.org

Department of Electrical and Computer Engineering The University of British Columbia Vancouver, BC, Canada V6T 1Z4 [17] K. Xu, M. Gerla, and S. Bae, Effectiveness of RTS/CTS handshake in IEEE 802.11 based ad hoc networks, Ad Hoc Network Journal, vol. 1, no. 1, pp. 107123, 200324 [18] P. Ball Broadband Wireless Networks Unpublished Lecture notes Feb., 2010 Oxford Brookes, University, UK [19] S. Hethley. Essentials of Wireless Mesh Networking Cambridge Wireless Essentials Series, 2009 Published in the United States of America by Cambridge University Press, New York [20] B. Azzedine Algorithms and Protocols for Wireless and Mobile Ad Hoc Networks Wiley Series on Parallel and Distributed Computing, 2009 university of Ottawa, Ottawa Canada FRANCE. [21] D. Nandiraju et al Service differentiation in ieee 802.11s mesh networks: a dual queue strategy CDMC,

Department of Computer Science, University of Cincinnati. [22] G. Brar et al Computationally Efficient Scheduling with the Physical Interference Model for Throughput Improvement in Wireless Mesh Networks [23] S. Suresh Performance of 802.11b Mesh Network under Channel Interference for Wireless Internet 2004 University of Sydney, Australia [24] X. Wu, et al Analysis of Bottleneck Delay and Throughput in Wireless Mesh Networks 2006 State Key Laboratory for Novel Software Technology, Nanjing University, China [25] C. Gomes Optimal Design of Wireless Mesh Networks 2007 I3S(CNRS-UNSA)/INRIA MASCOTTE project, Sophia Antipolis,

524