Key Technology and Experimental Research in Wireless Mesh Networks

1 Introduction to WMN
The Wireless Mesh Network (WMN) was first introduced in the middle of 1990s. With the development of mobile telecommunication technologies, there are increasing demands for voice services of wireless communication networks and high-speed Internet access. Traditional wireless access networks have difficulties adapting to current flexible and changing situations because of insufficient bandwidth, Quality of Service (QoS) problem, as well as lack of uniform planning in wireless spectrum resource allocation and topology. The WMN is thus introduced and it provides a new approach for solving the problems the traditional networks encounter.

     The topologies of traditional wireless access networks are mainly Point-to-Point (P2P) and
Point-to-Multipoint (P2MP), as shown in Figure 1. For example, the topology of cellular networks is P2MP; the Wireless Local Area Networks (WLANs) fall into two types by architecture: centralized and self-organizing, which topologies are P2MP and P2P respectively. The topology of WMNs is Mesh, which is also called Multipoint-to-Multipoint (MP2MP). This network architecture enables WMNs to solve the problems in traditional wireless access networks.

     The cellular network often covers a relatively large area and mainly provides voice services, but its construction and maintenance costs are high. Besides, in the communication hot-spots, it can only provide a relatively low rate for multimedia transmission. The WLAN provides a high rate, thus suitable for multimedia data access services. In case the WLAN has P2MP architecture where there is a center, the coverage of Access Points (APs) is limited. To provide wireless access in a larger area, lots of APs have to be deployed, and the cost increases. As for Ad hoc network, the network connectivity depends on the nodes with high mobility; consequently, the network reliability decreases. In addition, Ad hoc network requires gateway nodes with high performance, and its charging and management modes are not clear.

     The WMN mainly consists of two types of network nodes: Mesh routers and Mesh clients. When the WMN acts as an access network, its Mesh clients can access the Internet by means of wireless multi-hops between neighboring nodes. The MP2MP architecture gives the WMN the following advantages:

• Self-configuration: The nodes are interconnected into single-hop or
  multi-hops through open wireless links, thus they can automatically form a network.
• Self-adjustment: As there are often several communication paths between nodes, any service can flexibly select a proper path (e.g. the shortest path, the path with minimal interference or the highest rate path) for transmission.
• Self-healing: When a node is found damaged or congestion occurs on a link, the network services can select other nodes or links; hence the network reliability is enhanced.
• Scalability: The network can easily add or delete nodes, adjust the network coverage as well as reduce system construction and management costs.
  

     Therefore, the WMN is quite suitable for use as the Internet’s last-mile wireless access scheme, which has already been mentioned in IEEE 802.11s, 802.15 and 802.16 standards. It will make an important part of
next-generation Wireless Fidelity (Wi-Fi) and Worldwide Interoperability for Microwave Access (WiMAX) networks.

     Currently, there are many kinds of wireless access networks. They adopt different wireless transmission technologies and standards and can be applied in different scenarios. For instance, the cellular network is suitable for voice communications; Wi-Fi network is suitable for broadband multimedia data access services in local area; and the wireless sensor network is suitable for data collection in environment detection. The future wireless network will be a ubiquitous network, allowing all kinds of networks to co-exist, either heterogeneous or homogeneous, and providing the users with access services anywhere and anytime. This is a development trend of wireless networks and a potential application of WMNs.

     Figure 2 illustrates the architecture of a WMN integrated with various wireless networks. The core equipment in the WMN is Mesh routers, which form the core network with dynamic, changing topology. Each Mesh router is configured with several standard wireless transmission interfaces, enabling different wireless networks to be interconnected, including cellular network, WiMAX network, WLAN, Ad hoc network and wireless sensor network. Meanwhile, the user terminals, which adopt several wireless transmission technologies, can access the core network anytime and anywhere via a wireless access network. In terms of configuration and maintenance costs, the WMN has advantages over traditional wired networks. Therefore, it can be used as an implementation scheme for future ubiquitous heterogeneous networks[2-5].

2 WMN Networking Technologies
In a wireless network, interferences between nodes arise due to the following reasons: limited bandwidth, complex time-varying characteristic of channels and open communication environment. To provide QoS services as well as enhance its link capacity and network transmission efficiency, the wireless network has to adopt effective network management and networking technologies. The networking technologies of WMNs mainly involve in these aspects: network configuration and deployment, power control, mobility management and access control, and routing protocol design.

2.1 Network Configuration and Deployment
To provide the WMN with good scalability, powerful error-tolerance, excellent self-adjustment capability, as well as large coverage and capacity, it is necessary to configure and deploy the network carefully.
First of all, the deployment of Mesh routers with less mobility should be planned to form the backbone. On one hand, the number of Mesh routers should be as small as possible to reduce cost, and there should be no blind areas of wireless signals. On the other hand, multiple paths should be provided in hot-spot areas to enable more accesses. The integration of Mesh routers with Multiple-Input Multiple-Output (MIMO) system and directional antenna can further enhance the network’s transmission capacity.

      Mesh routers can provide access to various heterogeneous networks. As a result, a problem that has to be addressed in the WMN is how to allocate wireless interfaces of Mesh routers to ensure the connectivity between networks. In addition, using multiple channels in the WMN can increase the network’s throughput, but there are also many urgent problems to be solved for multi-channel communications in the wireless, multi-hop environment. Reference [6] discusses the challenges in multi-channel WMNs, such as distributed service assignment between channels, coordination between channels, and support for broadcast; meanwhile, it suggests protocols for multi-channel WMNs and their implementation.

2.2 Power Control
In WMNs, although there is not a restraint on power supply of each Mesh router, power control is still necessary in order to ensure WMN connectivity, control the interference and improve the spectrum’s reuse. A proper transmit power can reduce the interference between wireless channel signals and increase the spectrum’s reuse.

      Unlike Mesh routers, Mesh clients are often mobile nodes with limited energy. Therefore, the power efficiency should be considered in designing network protocols, especially when there are IP phones or sensors acting as Mesh clients. For some WMN applications, power control can optimize the network’s connectivity and improve its performance.

2.3 Mobility Management and Access Control
The mobility management of WMNs covers two aspects: location management and mobile handover management. Location management is mainly used for location-based registration; while mobile handover management often proceeds as follows: initialize a handover, set up new connection, and control the data stream during the handover so as to provide the user with seamless connection. In cellular networks, mobility management is implemented by base stations, the mobile switching center and the location database together in a centralized way; while in Ad hoc networks, mobility management is closely combined with routing protocols and distributed schemes are often used. In WMNs, which are characterized by multi-hop and low mobility, the technologies of the above two kinds of networks should be adopted in mobility management.

      The purpose of access control is to ensure that the users, who use different wireless transmission technologies, effectively access the WMN. So far, how to control the access of new nodes to guarantee the QoS of existing nodes is still a problem, especially in hot-spots where the number of nodes is large and the load is heavy.

2.4 Routing Protocol Design
Like Ad hoc networks, WMNs realize network access via multi-hop relays, so the methods used in Ad hoc networks can be referred to in designing routing protocols for WMNs. However, there are differences between the two kinds of networks: First, WMNs are made of APs, Mesh routers and gateways, so their mobility is relatively low; second, most data services in WMNs are transmitted from user nodes to gateway nodes, but services in Ad hoc networks are mainly transmitted in a peer-to-peer way; third, the power consumption restraint in WMNs is less than in Ad hoc networks.

      When the WMN is used for network integration, it can be regarded as wireless Internet. The Internet’s routing protocols and technologies have become quite mature. Therefore, in studying routing protocols of WMNs, the Internet routing technologies are very helpful and can be taken advantage of, for example, the technology that adopts different routing protocols inside and outside the autonomous domain.

3 Routing Protocol Design for WMNs
WMNs are multi-hop networks with dynamic topologies, similar to Ad hoc networks. Hence, most routing protocols adopted in WMNs are based on those of Ad hoc networks. The routing protocols of Ad hoc networks fall into two kinds: location-aided and non-location-aided. The location-aided routing protocols need the Global Positioning System (GPS), while the non-location-aided routing protocols can be further divided into two types: flat routing protocols and hierarchical routing protocols. The flat routing protocols include on-demand routing protocols (such as Ad Hoc On-demand Distance Vector (AODV), Dynamic Source Routing (DSR) and Segment-by-Segment Routing (SSR)) and active routing protocols (such as Destination-Sequenced
Distance-Vector (DSDV), Wireless Routing Protocol (WRP), and Topology Broadcast based on Reverse-Path Forwarding (TBRPF)), while the hierarchical routing protocols include Clusterhead Gateway Switch Routing (CGSR) and Zone Routing Protocol (ZRP). In existing WMNs, the routing protocols based on TBRPF[7], DSR[8], DSDV[9] and AODV[10] have been used.

      However, WMNs have some distinct differences from Ad hoc networks, as shown in Table 1. The direct application of Ad hoc network routing protocols in WMNs can not optimize the WMNs for  maximum performance.
In designing WMN routing protocols, the following factors have to be taken into account:

      (1) Routing Selection Criteria
      Several parameters should be considered in routing selection, including

• Hop count: The number of wireless links between the source node and the destination node;
• Expected Transmission Count (ETX): The expected number of retransmissions resulted from media access collision, which is a commonly used parameter in wireless communications;
• Expected Transmission Time (ETT): A parameter more common than ETX, which takes into account the bandwidth of the channel;
• Round-trip Transmission Time (RTT): The time required for the packet to take a round trip between the source node and the destination node;
• Energy consumption: The sum of energy consumed by all nodes in one transmission route;
• Route stability: The stability and duration of a route.
  

      (2) Supported Network Scale 
      The network scale of a WMN is generally huge, so the routing protocols should support a large number of nodes. If the routing protocols of traditional Ad hoc networks are directly used, the route search process may take quite a long time, thus resulting in considerable costs.

      (3) Error-Tolerance
      The error-tolerance capability enables the network to reselect a route in case that Mesh clients move, wireless links are congested or Mesh routers fail to work.

      (4) Link Interference
      The neighboring nodes will interfere with each other when they send wireless signals, so the link with the least interference should be selected in routing selection so as to increase the system capacity.

      (5) Cross-Layer Design
      A cross-layer design approach should be used to combine the directional antenna and MIMO technologies in Physical Layer and some technologies in Link Data Layer, ensuring that protocols of different layers coordinate.

      The WMNs support two applications: wireless access and interconnection of wireless networks. Depending on the application scenarios, the adopted routing protocols should focus on different aspects.

      In wireless access application, the routing protocols must fully adapt themselves to both Mesh clients and Mesh routers, involving both user nodes, which are characterized by high mobility and restrained power consumption, and access/gateway nodes, which are low mobile and without power consumption restraint. In most of existing WMN routing protocols, Mesh clients and Mesh routers are treated equally, and their differences have not been taken seriously. It may attract great attention to study the routing protocols based on the differences of the two kinds of nodes. When the WMN is applied to be integrated with other networks, the wireless network consisting of user nodes can be regarded as an autonomous domain and can adopt the routing concept of the Internet. In this case, it is only required to solve the routing problem of the wireless core network, which consists of Mesh routers. Reference[11] compares the performances of four commonly-used Ad hoc network routing protocols (namely DSR, AODV, Optimized Link State Routing (OLSR) and DSDV) in Mesh core network in terms of routing overhead, packet transmission success ratio, and end-to-end delay. The results show that the on-demand routing protocols (i.e. DSR and AODV) perform better than the active routing protocols (i.e. OLSR and DSDV) because their routing overheads are relative small. However, as the mobility of WMN nodes is low, the routing expire time and routing cache time of the on-demand routing protocols have to be extended in order to avoid too many routing messages being exchanged, thus leading to increase in overhead.

      In WMN’s wireless access application, most services are peer-to-peer services from user nodes to gateway nodes, having the feature of burst. When the WMN is integrated with other networks, the services between Mesh routers come from one network, so the service streams are convergent, similar to Internet services[12]. For example, the service range and average packet size follow the 24-hour model, and most IP services are TCP services, most of which are webpage browsing and file transfer. Therefore, to ensure QoS, service types should also be taken into account in the routing protocol design.

      When the WMN is integrated with other networks, Mesh routers are configured with more than two wireless transceivers, so the characteristics of multiple wireless transceivers and multiple channels should be considered in designing the routing protocols. While in WMN’s wireless access application, there is not such requirement.

      In summary, the research on WMN routing technologies should take consideration of the characteristics and applications of WMNs and take advantage of the technologies of wireless Ad hoc networks. Only this way, the performance of WMNs can be improved.

4 Experimental Research onWMNs
There are two main approaches to study WMN networking technologies: simulation and experiment. In simulation-base research, some simulation tools, such as the Network Simulator 2 (NS-2) and OPNET-based simulation software, are used to create a WMN protocol model and service transmission model; then protocol design and network performance analysis are made on the simulation results. In experimental research, a practical WMN experimental network, i.e. WMN testbed, is created by abstracting WMN network structure into small network architecture, service model and key technologies, and then tests and performance analysis of the WMN are all conducted on the testbed. This second approach is now being adopted by many international research institutes.

      For example, Microsoft Research (MSR) has set up an 802.11-based WMN experimental platform, and Massachusetts Institute of Technology (MIT) has established a Roofnet test platform. The objective of these platforms is to enable the WMN to provide Internet access services. In the WMN platform of Microsoft Research, a Mesh connection layer is added between the Network layer and the Media Access Control (MAC) sublayer of the node, and a modified DSR protocol is used as the routing protocol. The Roofnet platform of MIT consists of about 20 nodes and each node is configured with an 802.11 wireless network adapter and an
omni-directional antenna, both of which work in the same channel. The operating system is Linux and the routing protocol is Switched Routing Control Channel (SRCC) protocol, which is similar to DSR. A Mesh client obtains its IP address dynamically with Dynamic Host Configuration Protocol (DHCP) and accesses the Internet via the gateway node. In the testbed proposed in Reference [13], each Mesh router is configured with two network adapters: one adopts Host AP technology, used for wireless access service; while the other adopts table-driven routing protocol, used to construct the backbone. In the WMN platform presented in Reference [14], the routing protocol for the backbone network is IPv6-based DSDV protocol, which provides access to the client with the automatic configuration function of IPv6. Currently, IEEE 802.11-based wireless transmission technologies have become mature, and related products are of low prices and have been widely used. As a result, it has been regarded feasible and generally accepted to use these technologies to construct WMN testbeds required for WMN networking technologies research.

      In order to study WMN networking technologies and WMN node implementation schemes, we construct a WMN test platform here. This platform provides both wireless and wired access services, for instance,
IEEE 802.11-based WLAN access (e.g. AP access (which has a center) and Ad hoc access) and IEEE 802.3-based wired access, shown in Figure 3.

      When there are large numbers of Mesh clients in a WMN, these clients can form a subnet of Internet, accessing the Mesh backbone via Mesh routers. In some areas, the number of Mesh clients may be small. In this case, the access network can be regarded as a segment and be connected to the Mesh backbone via Mesh bridges. A Mesh bridge is similar to an Ethernet Hub or switch in function. Besides, it is inexpensive and can effectively forward data frames between wireless segments. Therefore, we design our implementation scheme based on Mesh bridges.

      With Mesh bridges, access to different access nodes can be realized, including WLAN access, Ad hoc access and wired access, and protocols can be switched across different access modes. One Mesh bridge-based implementation scheme is configured as follows: an Intel x86 notebook Personal Computer (PC) installed with V2.6.18 Redhat Fedora Core 6 operating system, the RJ45 interface of this notebook PC used to connect the wired network, and a wireless network adapter added with Personal Computer Memory Card International Association (PCMCIA) bus for wireless access. The model of this wireless network adapter is DWL650, on which Intersil’s Prims 2/2.5/3 chip set and Host AP driver are used. This wireless adapter supports Host AP mode and can provide IEEE 802.11 access function, just like APs.

      Based on the abovementioned hardware and software, we construct a minimal WMN experimental system, as shown in Figure 4. The protocol translation function of Mesh bridges is performed with Libpcap and
Libnet-based technologies. Among them, Libpcap-based technology is for capturing data frames, while
Libnet-based technology is for transmitting data frames. By capturing the data frames on the transmission links and using the Layer 2 (link layer) forwarding technology, the Mesh bridge forwards data between segments and converts protocol frame formats. As the system adopts Layer 2 forwarding technology rather than Layer 3 IP forwarding technology, used for non-Mesh routers, data forwarding is sped up and the forwarding time is shortened.

      As shown in Figure 4, the experimental system consists of two main parts: two Mesh bridges and three Mesh clients. Each Mesh bridge is configured with two communication interfaces: one is for Mesh client access (either Ethernet access or WLAN access), and the other is for connection to Mesh backbone. The latter adopts AODV routing protocol of Ad hoc network, acting as a switch node of the Mesh backbone. Mesh clients fall into two kinds by their access modes: WLAN access and Ethernet access. IEEE 802.11b supports 11 channels, three of which are not overlapped. Among these channels, Channel 1 and Channel 6 are allocated to the Mesh bridge, and the Ad hoc connection between Mesh bridges uses channel 11, so all wireless channels will not interfere with each other.

      In this system, data communications between any two clients can be realized; hence this system can be used to further study of Layer 2 routing and switching algorithms. Reference [15] proposes a Multi-hop Address Resolution Protocol (MARP) scheme, which completes the routing request and reply processes by extending Address Resolution Protocol (ARP). The next goal of this experimental system is to design an effective Layer 2 switching algorithm and to verify its performance with experiments.

5 Conclusion
This paper introduces the architecture and makeup of the WMN and analyzes the wireless routing protocols and network management technologies for WMNs, pointing out the key technologies and main challenges involved. Moreover, it constructs a WMN experimental platform and tests its performance. Experiments show that the WMN is a feasible wireless access network architecture and a networking technology with good application prospect.

References
[1] WAHARTE S, BOUTAB R, IRAQI Y, et al. Routing protocols in wireless mesh networks: challenges and design considerations [J]. Multimedia Tools and Applications, 2006, 29 (3): 285-303.
[2] AKYILDIZ I F, WANG Xudong, WANG Weilin. Wireless mesh networks: a survey [J]. Computer Networks, 2005, 47 (4): 445-487.
[3] ESTRIN D, CULLER D, PISTER K, et al. Connecting the physical world with pervasive networks [J]. IEEE Pervasive Computing, 2002, 1 (1): 59-69.
[4] HIERTZ G R, MAX S, WEISS E, et al. Mesh technology enabling ubiquitous wireless networks [C]//Proceedings of 2nd Annual International Wireless Internet Conference, Aug 2-5, 2006, Boston, MA, USA. New York, NY, USA: ACM. 2006.
[5] 彭木根, 姜涌, 王文博. 无线数字家庭网络泛在接入技术 [J]. 中兴通讯技术, 2008,12 (4): 41-46.
[6] KYASANUR P, JUNGMIN S, CHEREDDI C, et al. Multichannel mesh networks: challenges and protocols [J]. IEEE Wireless Communications, 2006, 13 (2): 30-36.
[7] Firetide website [EB/OL]. http://www.firetide.com.
[8] Networking Research Group. Microsoft Research. [EB/OL]. http://research.microsoft.com/mesh/.
[9] MIT Roofnet [EB/OL].
http://pdos.csail.mit.edu/roofnet/.
[10] PIRZADA A, PORTMANN  M, INDULSKA J. Evaluation of multi-radio extensions to AODV for wireless mesh networks [C]//Proceedings of 4th ACM International Workshop on Mobility Management and Wireless Access (MobiWac’2006), Oct 2, 2006, Malaga, Spain. New York, NY, USA: ACM, 2006: 45-51.
[11] ASHRF U, JUANOLE G, ABDELLATIF S. Evaluating routing protocols for the wireless mesh backbone [C]// Proceedings of 3rd IEEE International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob’07), Oct 8-10, 2007, New York, NY, USA. 2007: 40-47.
[12] THOMPSON K, MILLER G J, WILDER R. Wide-area Internet traffic patterns and characteristics [J]. IEEE Network, 1997, 11 (6): 10-23.
[13] SONG Heecheol, KIM Bong Chan, LEE Jae Young, et al. IEEE 802.11-based Wireless Mesh Network Testbed [C]//Proceedings of 16th IST Mobile and Wireless Communications Summit, Jul 1-5, 2007, Budapest, Hungary. 2007: 1-5.
[14] IANNONE L, KABASSANOV K, FDIDA S. Wireless mesh network testbed demo [C]//Proceedings of IEEE International Conference on Mobile Ad hoc and Sensor Systems (MASS’06), Oct 9-12, 2006, Vancouver, Canada. Piscataway, NJ, USA: IEEE, 2006: 582-585.
[15] LEE Sung Hee, KO Young Bae. An efficient
multi-hop ARP scheme for wireless LAN based mesh networks [C]// Proceedings of 1st Workshop on Operator-Assisted (Wireless Mesh) Community Networks, Sep 18-19, 2006, Berlin, Germany. Piscataway, NJ, USA: IEEE, 2006: 1-6.

[Abstract] With the increasing popularity and rising demand for high-rate wireless Internet access, traditional wireless access networks such as cellular network and Wireless Local Area Network (WLAN) are facing some challenges. The Wireless Mesh Network (WMN) is emerging as a flexible and low-cost alternative to provide multi-hop communications, supporting applications such as last-mile Internet delivery. The WMN has also become a promising technology in the merging of wireless networks. Some key technologies for the WMN networking, including network configuration, power control, mobility management, access control, and routing protocols, are analyzed; the routing protocol design in WMN and mobile Ad hoc network are compared. An example of WMN testbed based on WLAN and second-layer switching technology is given.