Managing Wireless Mesh Networks- Analysis and Proposals

Djohara Benyamina*, Abdelhamid Hafid*, Michel Gendreau*, Nasreddine Hallam+ *Networking lab (LRC), School of Operational Research, Department of Art & Sciences

University of Montreal, Montreal, Canada + University of Nottingham, School of computer science and IT

benyamid@iro.umontreal.ca

Abstract

A mesh network allows nodes or access points to communicate with each other without being routed through a central switch point, eliminating centralized failure and providing self-healing & self-organization. Although decisions on traffic are made locally, the system can be managed globally. In this paper, a preliminary study on Wireless Mesh Networks (WMNs) management is presented, pointing out the different managements functions required to maintain the network reliability. A decomposition of WMNs management functions into key areas is presented and analyzed. The state-of-the-art management protocols for WMNs are explored and classified according to the proposed key areas.

1. Introduction

Wireless broadband networks are being increasingly deployed in a multi-hop wireless mesh network (WMN) configuration. In WMNs, the access points (mesh routers) are rarely mobile and may not have power constraints. In addition these networks behave almost like wired networks in having infrequent topology changes, limited node failures etc.

Although WMNs may be self-organizing and self healing, node addition and maintenance are still rare events. Furthermore, since each mesh router may aggregate traffic flows for a large number of mobile clients, the aggregate traffic load of each mesh router changes very rarely. In infrastructure WMNs (IWMNs), some mesh routers are also equipped with a gateway capability through which they interface with the wired network. In such networks, traffic is mainly routed by the WMN wireless backbone between the mesh clients and the wired Internet and goes through the gateway

nodes. Figure 1 presents a hybrid architecture where both mesh clients and mesh routers constitute network.

Such networks (WMNs) are deployed in different scenarios: home networking, enterprises, universities, and public places at a remarkable rate. However, users experience a number of problems such as intermittent connectivity, poor performance, lack of coverage, and authentication failures. Some industry-driven solutions [18][15], combining existing cellular, ad hoc, and sensor network concepts are offered. However, there is a need for an increased focus on the unique challenges and opportunities that a WMN presents, and based on these qualities, new management mechanisms, specially tailored for the WMN need to be developed.

This paper is structured as follows. Section 2 is a general introduction to network management. The presentation of management of wireless mesh networks in particular is presented in section 3. In this section the critical factors influencing network performance are also shown. The major differences between ad hoc and mesh networks are shown in section 4. Then in section 5, the different WMNs management functions are defined and classified among four key areas. Section 6 presents some considerations to be taken as a result of our initial study on WMNS management. Finely, we conclude our work in section 7.

Figure 1. Hybrid WMN ( Figure taken from [17])

Third IEEE International Conference onWireless and Mobile Computing, Networking and Communications (WiMob 2007)0-7695-2889-9/07 $25.00 © 2007

2. Network Management: Overview

A good definition of network management can be found in [19]: “network management is a process of controlling a complex data network so as to maximize its efficiency and productivity”.

To accomplish this process, network management can be functionally divided into five areas defined by (ISO) [19, 20]: fault management, configuration management, performance management, security management and accounting management.

Based upon the information collection and communication strategy, there are three types of network management architectures: Centralized architecture, Distributed architecture and Hierarchical architecture. The latest uses intermediate managers to distribute the management tasks. Based on the characteristics of the architectures mentioned above, the centralized architecture has to be omitted when dealing with a large telecommunication networks. The main reason is that a network partition occurs because of some kind of failure, the disconnected portion from the manager is left without any management functionality.

However a higher reliability and efficiency is achieved with a distributed architecture as well as lower overhead both on communication and computation resources. This approach has been adopted by many management systems such as Telecommunication Management Network (TMN) [21] and Management model for ATM networks. The hierarchical architecture seems to be more efficient than the other, and was initially proposed in[15]. 3. Management in Wireless Mesh Network By definition of Wireless Mesh network architectures (Infrastructure, Mesh clients, Hybrid) and their features (self tuning, self healing), it has become inefficient to use the five areas provided by ISO to develop a management prototype framework for WMNs. So we need first to define the different areas from where a given WMN could be managed. 3.1. Critical factors influencing network performance

The management problems constitute mesh technology’s biggest technical hurdles. Before a network is designed, deployed, and operated, factors that critically influence its performance need to be considered. So what are the critical factors influencing network performance?

For WMNs, the critical factors can be summarized as follows: Radio Techniques: Currently many approaches have been proposed to increase capacity and flexibility of wireless systems. Directional and smart antennas[22][25], MIMO systems[24][26], and multi-radio/multi-channel systems[23][27] are typical examples for such approaches. Scalability: It has been shown and proved [13] that when the size of network (multi-hop network) increases, the network performance degrades significantly. Mesh connectivity: It is a critical requirement for Mac and routing protocols. Topology-aware MAC and routing protocols can significantly improve the performance of WMNs. Broadband and QoS: More performance metrics such as delay jitter, aggregate and per node throughput, and packet loss ratios, must be considered by communication protocols. Compatibility and inter-operability: Integration of WMNs with other wireless networks requires certain mesh routers to have the capability of inter-operation among heterogeneous wireless networks. Security: secure MAC and routing protocols, intrusion detection, and security monitoring need to be developed. 3.2. Faults in a wireless network

Network problems such as connectivity, performance, security, and authentication are derived from interviews and discussions conducted between network specialist authors [11] and network administrators of Microsoft’s IT department. The connectivity problem concern the inconsistent or lack of network connectivity is due mainly because of weak RF signal. Degraded performance (low throughput or high latency) is due to congestion,, RF/multi-path/large co-channel interference. Many contributions in this context have been proposed (see for example [2], [10], [11]) and will be analysed later in the next session. Although, IEEE 802.1x offer security solutions to their networks clients. However, IT managers are still dealing with an important wireless network security problem known as “Rogue AP Problem”[28][29][30] due to unauthorized AP connected to an Ethernet tap of the corporate network. Authentication failure is due to missing or expired certificates.

4. Mesh Networks Vs. ad hoc Networks

There are various management protocols for ad hoc network SNMP[30], CMIP[19][30], ANMP[15] and Guerrila[16]. So why not to use an ad hoc management protocol for wireless mesh network?

Although a WMN is similar in concept to a mobile ad hoc network, there are some important differences

Third IEEE International Conference onWireless and Mobile Computing, Networking and Communications (WiMob 2007)0-7695-2889-9/07 $25.00 © 2007

between the two. A detailed comparison can be found in [7]. Firstly, nodes in a WMN are fixed ( not mobile). Topology changes are therefore infrequent. Secondly, the traffic characteristics, being aggregated from a large number of traffic flows, and not changing very frequently. Thirdly, the traffic distribution in a WMN is typically skewed as most of the user traffic is directed to/from a wired network. Finally, WMNs require proactive discovery of paths to reduce packet delays, whereas most mobile ad hoc networks use reactive routing strategies. Therefore, we propose another approach of classifying WMN management functions. This is detailed in the rest of the paper. 5. Management Functions in WMNs

In this paper we propose a different decomposition approach as depicted in figure2 (a)(b).

Figure 2. (a) A proposed key areas of managing a WMN. (b) The different types of managements

proposed by I.F.Akyildiz[17]. The power management when dealing with the power

consumption and not the power transmission (defined by the range of the transmission) is classified as a mobility-driven management where mesh clients may expect protocols to be power efficient. Admission control and QofS routing is replacing the power management, more precisely for power transmission. Admission Control and QofS routing are disjoint sub-modules, which can be linked in a joined area management as done in [4][6]. 5.1 Network Monitoring

In order to continuously monitor the network performance, statistics of mesh nodes, especially mesh routers, need to be reported to one or several servers

according to the management architecture used, (see section 1.1). These statistical data are analysed by data processing algorithms in the performance monitoring software on the server, eventually determine potential abnormality in which case the server reacts to take responses.

Since a performance degradation is in fact another way to show the existence of faults, as described in 3.2, we found it necessary to classify contributions in this area into two sub-classes. The first sub-class, (say “class1”), gathers all performance related works. In this category, authors generally, propose implicitly a fault recovery technique within their routing algorithms. The second class, class2, encompasses all directly dedicated fault detection and/or diagnosis contributions.

We can broadly classify the researchers’ works related to wireless network management into three areas: 1) protocols for network management,2) mechanisms for detecting and correcting routing and MAC misbehavior, 3) general fault management. Generally, class 2.3 groups market commercial products such as AirWave[29], Symbol’s wireless network management system (WNMS) [32], etc. 5.1.1. Contributions in network monitoring. Classe 1 -Ashish et al.[7] propose a channel assignment and routing techniques in a multi channel WMN. The channels are dynamically assigned through a distributed algorithm that utilizes only local traffic load information. This version eliminates the need of separate control interface and incorporates prioritized channel assignment to emulate a logical fat tree structure. It also supports fast failure recovery.

In the same context, a very interesting metric for routing in multi-radio, multi-hop wireless networks was introduced by Draves et al. [1]. The goal of the metric is to choose a high-throughtput path between a source and a destination. The metric named WCETT is defined as a combinasion of weights assigned to individual links based on the Expected Transmission Time (ETT) which is a function of the loss rate and the bandwidth of the link. In their work, as a result of their conducted experiments, they conclude that, a path that is made up of hops on different channels is better than a path where all the hops are on the same channel (interference problem consequence).

A novel throughput optimization technique in Multi-radio WMNs [5] deserve to be presented in this paper. The authors mathematically formulate the joint channel assignment and routing problem, taking into account the interference constraints. Their mathematical model is then used to develop a solution that optimizes the overall network throughput subject to fairness constraints on allocation of scarce wireless capacity among mobile clients. The initial mathematical model

Mobility management Admission

Control

QofS Routing

Network Monitoring

- peerToPeer - QofS routing Based on power consumption

Security Management

Power Management

Network Monitoring

Security Management

Mobility Management

Third IEEE International Conference onWireless and Mobile Computing, Networking and Communications (WiMob 2007)0-7695-2889-9/07 $25.00 © 2007

is built on the basis of link flow scheduling necessary and sufficient conditions. Thus to maximise the throughput for each node (as defined in their graph model), many subproblems need to be solved (traffic rate, feasible channel assignment, feasible schedule S within a schedule period T).

Analysis and proposition. As the interference free edge scheduling sub-problem itself is NP-hard, the authors present an approximation algorithm for the overall problem, hence more other steps are required to refine the solution of the relaxed problem. They used a linear program (PL) to find a flow that maximizes the throughput. The same PL is solved twice with different objective functions. The resolution technique they use is nothing but an instance of the well known aggregated objective resolution. Some experiment results [43] show many drawbacks of this technique. These problems stem naturally from the aggregate approach which is vulnerable to concave objective landscapes. Basically when the landscape of the single objective resulted from aggregating the many objectives, is not convex, then the image of the solutions located on those concave regions might very well be overlooked ( see figure 3.)

We propose to - Define two independent objective functions (given by (9) and (10) in[5]) and associate them to the set of constraints defined by authors’ PL . - Solve the final PL using swarm based optimisation techniques, which has been proved to be a robust and good resolution tool for NP-hard problems [42][44].

Figure 3. Image of the solutions located on the concave regions might be overlooked.

Class2- A new direction of research on fault diagnosis in WMNs is proposed in [10]. The authors present a diagnosis system that employs trace-driven simulations to detect faults and perform root cause analysis. They use the approach to diagnose performance problems caused by packet dropping, link congestion, external noise, and MAC misbehaviour. However, the tool presented does not detect link attacks and can report and detect only one fault at a time which is not the usual case in the real world (Classified in class 2.2).

Sergio et al.[2] propose a watchdog mechanism to detect network unreliability problems stemming from

selfish nodes. The basic idea is to have watchdog nodes observe their neighbours and determine if they are forwarding traffic as expected; included in classe2.2. This approach for detecting routing anomalies has been further refined by others as well (see [34][35]). The limitation of this approach is that it relies on the presence of nearby clients for diagnosing some of the wireless faults. Adya et al.[11] present an architecture and techniques for diagnosing faults in IEEE802.11 infrastructure Networks, where they propose a novel technique which enables bootstrapping and fault diagnosis of disconnected clients, they also present an approach for detecting unauthorized access point, what we called in 3.3 Rogue APs. Classified in class 2.1 5.2. Mobility Management

With reduced coverage areas compared to a cellular network, the degree of mobility increases dramatically, requiring novel solutions to ensure that mobility-related overhead does not overwhelm the network. Conventional problems persist, including managing handoffs, location, and changing network demands. However, the mesh nature of the network also provides a new level of interconnection to be exploited in quickly responding to local handoffs. When the mesh clients constantly roam across different mesh routers, the need to define the topology of the WMN is necessary to enhance the performance of MAC and routing protocols. A distributed mobility management architecture is a preferred solution for WMNs as stated in [17]. Another interesting issue in mobility management is the constraint on power consumption. Therefore, a need to develop a multi-layer management schemes is still an interesting research area.

Mobility management mechanisms for various wireless networks like 802.11 mesh networks and WIFI/WIMAX relay networks were proposed [9]. An important point was discussed in this paper, figuring out enhanced P2P mobility schemes for achieving session continuity and seamless handover.

A multi-layer optimization technique addressing the two sets of challenges; Optimization of multicast routing and power control for a WMN was proposed[8]. A general primal-dual algorithm to solve the throughput maximization problem is implemented in distributed manner. 5.3.Admission Control & QoS Routing

The scarcity of wireless resources [18] suggests that demand will always outpace availability. Therefore, in order to ensure that QoS guarantees can be upheld, admission control must be used to limit the

Third IEEE International Conference onWireless and Mobile Computing, Networking and Communications (WiMob 2007)0-7695-2889-9/07 $25.00 © 2007

volume of traffic within the network. However, fluctuating wireless resources and the presence of congestion on wireless mesh links add to an already difficult research problem.

The mesh provides a high connectivity which can be exploited to allow for the flexible and dynamic allocation of resources within the network.

Admission control depends on precise estimates of bandwidth available in the network and the bandwidth required by a new flow. Estimating these parameters in wireless networks (more precisely for ad hoc networks) is challenging due to the shared and open nature of the wireless channel. Some QoS solutions have been proposed for wireless ad hoc networks. Where few of them [5][38][39] do not consider the fact that nodes could interfer with each other even though they may not communicate directly. The contention-aware admission control protocol where the concept of neighbourhood available bandwidth was introduced in [37],[4] and [6].

5.4. Security Management

There are implicit trust assumptions assumed in a mesh network (e.g., the nodes belong to the same administrative and security domain) unlike the random and arbitrary collection of nodes in an ad-hoc network.

Security in terms of authentication and authorization is not a big issue for wireless LAN, some of the standards activities for security, focus on inter-AP security controls, where client access uses standard WPA2/802.11i authentication and encryption.

Another security issue comes from the fact that the Multi APs (MAPs) are typically installed in public areas where physical security is low, so the opportunities for rogue APs (defined in 3.3) and intruders to access the network is far greater than in a closed, private office space. Others threats and vulnerabilities in Wireless Mesh Networks are either related to routing protocols or WiFi access such as Black-hole, Grey-hole, Worm-hole, Route error injection and Denial-of-service, Theft-of-service attacks.

The four first attacks are unique to wireless mesh networks [12]. Black-hole, grey-hole and worm-hole require packet injection with a specialized knowledge of the routing protocol, whereas route error injection does not require such detailed knowledge.

To enhance security of WMNs, two strategies need to be adopted. Either to embed security mechanism into network protocols such as secure routing and Mac protocols or develop security monitoring and response systems to detects attacks, monitor service disruption, and respond quickly to attacks. Some secure protocols propositions [40][41] are one layer problem solution.

Thus defending attacks is limited, because problems in other layers cannot be solved. 6. Some considerations

Even having the capability of self-organization in WMNs that reduces the complexity of network deployment and maintenance, and thus, requires minimal upfront investment. A new management mechanisms need to be developed, due to specific features of WMNs.

When considering mesh clients, a distributed mobility management is a preferred solution for WMNs, if compared to centralized. ( see 2.1). Another important point is that, mobility management is closely related to multiple layers of networks protocols. So a multi-layer mobility management scheme need to be developed for WMNs. For a security monitoring system, a cross layer framework also needs to be developed. However how to design and implement a practical security monitoring system, including cross-layer secure network protocols and various intrusion detection algorithms is still an open issue. 7. Conclusion

Wireless Mesh Networks have emerged as an industry-driven initiative, combining several existing technologies and concepts from cellular, ad hoc, and sensor networks. The result is a new network with a unique characteristics that create new challenges and opportunities.

In this paper, A detailed study was conducted, to show the different ways, from where, such network would be managed. To put into place a management framework, four key areas of management, have been defined. Many contributions in the context were explored and the main drawbacks of some of them are shown and new ideas are suggested.

Some interesting considerations, to develop an efficient management protocol for WMNs are highlighted. Intelligent management mechanisms will enable the evolution of WMNs into the next generation of wireless access networks. References [1] Richar Draves, Jitendra Padhye, and Brian Zill, Routing in Multi-Radio, Multi-Hop Wireless Mesh Networks, MobiCom’04, Sept. 26-Oct. Copyright 2004 ACM. [2] Sergio Martini et al. Mitigating Routing misbehaviour in Mobile Ad Hoc Networks. Intl. conf. on Mobile computing and networking, Boston, United States (255–265) 2000.

Third IEEE International Conference onWireless and Mobile Computing, Networking and Communications (WiMob 2007)0-7695-2889-9/07 $25.00 © 2007

[3] Payman Arashahi et al., Adaptive Routing in Wireless Communication Networks using Swarm Intelligence, 9th AIAA Int.., 17-20 April 2001, Toulouse, France. [4] Lin Luo, et al , A QoS Routing and Admission Control Scheme for 802.11 Ad Hoc Networks, DIWANS’06, Sept.25. Copyright 2006 ACM. [5] Mansoor Alicherry et al. , Joint Channel Assignement and Routing for Throughput Optimization in Multi-radio Wireless Mesh Networks, MobiCom’05, Aug. 28-Sept. 2, Germany. [6] Yuxia Lin et. An Admission Control Algorithm for Multi-hop 802.11e based WLANs, QShine’06, ACM.Aug. 7-9. [7] Ashish Raniwala, Tzi-cker Chiuch, Architecture and Algorithms for an IEEE 802.11-based Multi-channel Wireless Mesh Network, INFOCOM 2005. Proceedings IEEE, P.13-17,Volume3(2005). [8] Jun Yuan et al. , A cross-Layer Optimization Framework for Multihop Multicast in Wireless Mesh Networks, Wicon, Budapest, Jul. Copyright IEEE 2005. [9] Wei-Peng Chen et al, QoS Management and Peer-to-Peer Mobility in Fixed-Mobile Convergence, FUJITSU Sci. Tech. j., 42,4,p.535-546 (Oct. 2006). [10] Lili Qiu, et al. Troubleshooting Wireless Mesh Networks, ACM SIGCOMM Volume 36, p.17-28, Issue 5 (October 2006) ISSN:0146-4833. [11] Atul Adya, Paramvir Bahl, Ranveer Chandre, and Lili Qiu. Architecture and Techniques for Diagnosing Faults in IEEE 802.11 Infrastructure Networks. MobiCom’04, sept. 26-Oct. Copyright 2004 ACM [12] A.Gerkis, J. Purcell. A survey of Wireless Mesh Networking Security Technology and Threats. Gold paper. SANS Institute , 2006. [13] White paper. Solving the Wireless Mesh Multi-hop Dilemma. StrixSystems, 2005. [14] Mesh Casts Wider Network. http:// www.eetimes.com/news/latest/showArticle.jhtml [15] W. Chen, N. Jain and S. Singh, ANMP: Ad hoc network network management protocol, IEEE Journal on Selected Areas in Communications 17(8) (August 1999) 1506--1531. [16] Chien-Chung Shen et al, The Guerrilla Management Archtecture for Ah hoc Networks, [17] Ian F. Akyildiz et al., Wireless mesh networks : a survey. Copyright 2004 Elsevier. [18]Resource Management in Wireless Mesh Networks. http://bcr2.uwaterloo.ca/~mesh/ [19] Allen Leinwand and Karen Fang, Network Management A Practical Perspective, Addison Wesley, 1993. [20] C.S. Raghavendra and Suresh Singh, : PAMAS- Power Aware Multi-Access protocol with Signaling for Ad Hoc Networks”, Technical Repsrt TR-9801, Department of Cs, Univ. of South Carolina, Columbia, March 24, 1998. [21] Mohsen Kahani and H. W. Peter Beadle, “Decentralized approach for Network Management”, ACM SIGCOM Computer Communication Review, January 1997, pp. 36-47. [22] R. Ramanathan, On the performance of ad hoc networks with beamforming antennas, in: ACM (MOB/HOC), oct 2001, pp. 95-105. [23] J. So, N. Vaidya, Multi-channel MAC for ad hoc networks: handling multi-channel hidden terminals using a single transceiver, in: ACM (MOB/HOC), May 2004, pp. 222-233.

[24] W. Xiang, T. Pratt, X. Wang, A software radio testbed for two-receiver space time coding wireless LAN, IEEE Comminication Magazine 42(6) (2004) S20-S28. [25] A. Spyropoulos, C. S. Raghavendra, Asymptotic capacity bounds for ad hoc networks cases, in: IEEE (GLOBECOM), 2003, pp. 1216-1220. [26] K. Sundaresan, R. Sivakuman, M.A. Ingran, T.-Y. Chang, A fair medium Access Control Protocol for ad hoc networks with MIMO links, in: IEEE (IFOCOM) 2004, pp. 2559-2570. [27] A. Adya, P. Bahl, J. Padhye, A. Wolman, L. Zhou, A Multi-radio unification protocol for IEEE 802.11 wireless networks, in ( BROADNETS), 2004. [28] AirDefense. Wireless LAN Security. http://airdefense.net [29] AirWave. AirMagnet Distributed System. http://airmagnet.com [30] William Stallings, SNMP, SNMPv2, and CMIP: The practical guide to network management standards, 1st edition, 1993 Addison-wesley Publication Compagny, Inc. [31] Cisco. CiscoWorks Wireless LAN Solution Engine. http://Cisco.com. [32] SpectrumSoft: Wireless network management System, Symbol Technologies Inc. http//www.symbol.com [33] R. Mahajan et al., User-level Internet Path Diagnosis. In ACM SOSP, Bolton Landing, NY, Oct 2003. [34] B. Awerbuch, D. Holmer, and H. Rubens. Provably Secure Competitive Routing against Proactive Byzantine Adversaries via Reinforcement Learning. In JHU Tech. Report, Version 1, may 2003. [35] S. Buchegger and J. Le Boudec. The Effect of Rumor Spreading in Reputation Systems for Mobile Ad-Hoc Networks. In Proceeding of WiOpt, France, March 2003. [36] G. Ahn, et al., `SWAN: Service Differentiation in Atateless Wireless Ad Hoc Networks”, Proc. IEEE INFOCOM`2002, New York, June 2002. [37] Y. Yang and R. Kravets. Contention-Aware Admission Control for Ad Hoc Networks, Tech Report 2003-2337. University of Illinois at Urbana – Champaign, April 2003. [38] Y.Dong, D. Makrakis and T. Sullivan, Effective Admission Control in Multiple Mobile Ad Hoc Networks, IEEE Proc. ICCT 2003. [39] M. G. Barry, et al., Distributed Control Algorithm for Service Differentiation in Wireless Packet Networks, emphProc, IEEE INFOCOM 2001, New York, april 2001. [40] Y. HU, et al., A secure on demand routing protocol for ad hoc networks, ACM (MOBICOM), Sept. 2002. [41] M. Zapata, N. Asokan, Securing ad hoc routing protocols, ACM workshop on wireless security (Wi Sc), Sept. 2002, pp.1-10. [42] Coello Coello, Carlos A. (1999). A Comprehensive Survey of Evolutionary-Based Multiobjective 0ptimization Techniques. Knowledge and Information Systems, Vol. 1, No.3,pp.269-308. [43] Das, I. and Dennis, J. (1997). A closer look at drawbacks of minimizing weighted sums of objectives for Pareto set generation in multicriteria optimization problems. Structural Optimization, vol. 14, pp. 63-69. [44] Parsopoulos, K.E., Vrahatis, M.N. (2002). Particle Swarm Optimization Method in Multiobjective Problems. Proceedings of the 2002 ACM (SAC 2002), pp. 603-607.

Third IEEE International Conference onWireless and Mobile Computing, Networking and Communications (WiMob 2007)0-7695-2889-9/07 $25.00 © 2007