Anand Patil. A wireless mesh network is a self-organized backbone wireless network that can

Transcription

1 Comprehensive survey of handoff management challenges in wireless mesh networks and the evaluation of MPLS-based solution for providing an efficient transport Anand Patil 1. Research Topic A wireless mesh network is a self-organized backbone wireless network that can dynamically maintain connectivity between its wireless nodes. The mesh here refers to an interconnected group of mesh routers that help maintain connectivity throughout the mesh network. A mobile terminal which is connected to a mesh network can maintain the network connectivity as it moves from one Access Point to another within the mesh network. The traditional wireless access methods such as Wi-Fi networks, WLAN, Bluetooth, and cellular networks use single hop communication; for this reason they are connected to an Access Point which connects directly to a wired backbone network. Expanding the coverage of traditional wireless access methods is expensive because the wired backbone infrastructure which provides the backhaul service requires new cables to be laid. In many situations laying new cables is not feasible for practical and economical reasons. In contrast to single hop communication, the wireless mesh networks (WMN) use multi-hop communication. In the wireless mesh networks each Access Point (AP) is wirelessly connected to mesh routers. The mesh routers form the wireless backbone for WMN networks by routing packets. In wireless mesh networks extending the wireless coverage to larger areas is simple and cost-effective. This is the reason that service providers such as Aruba Networks and Cisco are deploying next generation wireless mesh 1

2 networks because they are becoming the preferred way to deliver Video, Data and Voice in outdoor environments. A wireless mesh can deliver the same network capacity, reliability and security that were once reserved for wired networks but with the flexibility of wireless. With today s state-of-the-art solutions, municipalities, public safety agencies, port authorities, and industrial organizations can rely on mesh networks to provide essential connectivity to their workers and constituents (Aruba Networks, 2011). When the city of Austin, TX decided to offer cheaper way of providing wireless access to its public on-the-go, the Austin city officials decided to build a wireless mesh network in the city that would link the public places. The city of Austin successfully installed wireless mesh solution with the help of Cisco s mesh infrastructure (Cisco, ). Handoff management which is part of the mobility solution for mesh networks is a determining factor for successful deployments of wireless mesh network. WMN design poses challenges in terms of the mobile device changing its attachment to the network across subnets or IP domains. Wireless service providers such as Cisco and Aruba Networks are faced with the problem of providing a consistent mobility management solution in a wireless mesh network. Aruba Networks has adopted High-speed outdoor roaming which is a modified version of the Mobile IP mobility solution. Aruba network s proprietary MobileMatrix technology for fast roaming of mobile terminals across IP subnet is an example of a mobility solution (Aruba, 20011, pp-9). This paper will look at few of these vendor implementation of mobility management solution. Two solutions for mobility management namely mobile IP and MPLSbased modification of Micro-mobile IP are described later in the paper. This paper analyzes handoff and routing needs of the WMN networks, examines the MPLS application micro mobility solutions, thus evaluating the suitability of MPLS as an efficient transport 2

3 solution. The paper also looks at the current status of IETF and IEEE standards in terminal mobility area. 2. Research Question What are the mobility management challenges unique to the wireless mesh networks and how have the standardization efforts in this area addressed the WMN deployment issues? Can the MPLS routing features help achieve efficient handoffs for wireless mesh networks? 3. Background A wireless mesh network is basically a collection of fixed wireless nodes, most of the time consisting of regular wireless routers running adapted software. Its main goal is to provide an inexpensive and easily deployable wireless backhaul that will connect distant LANs or WLANs (Carrano, et al., 2011, p. 54). As opposed to wireless mesh network, the wired backhaul refers to a wired backbone network capable of routing between nodes. Typically wired backbones are an internet network composed of access routers, edge routers, and gateways interconnected with each other using backbone wired network. The Mobile Terminals (MT) accesses the network using end devices. These end devices are connected to part of networks which are usually lower capacity network referred to as access networks. Access networks are the network which MT are attached to. WLAN, for high-speed wireless data network, cellular network, for voice and data connectivity, satellite communications for wireless access to commercial applications, and Wi-Fi hotspot networks in public places for wireless network connectivity are all examples of access networks (Akyildiz et al., 2004). A WMN has to be able to integrate with these heterogeneous access networks with acceptable handoff latency. 3

4 The term mobility can refer to either nomadic(terminal) mobility, vehicular mobility, service mobility or session mobility. Service mobility refers to the ability to maintain sessions and obtain services without any interruption while user changes terminals and session mobility allows the user to maintain a session while he/she changes terminals such as a laptop to a PDA (Mohammad et al, 2009). Only the nomadic/terminal mobility is in the scope of this paper. The nomadic mobility refers to maintaining the identity of the terminal as it moves from one location to another. So as the mobile terminal moves from one location to another location, its peer devices need to be able to continue communication with it using the terminal s known identity irrespective of its location (Schiller & Voisard, 2004, p. 213). Nomadic mobility is also referred to as terminal mobility. Mohammad et al. (2009) define terminal mobility as The ability of a terminal, while in motion, to access telecommunication services from different locations, and the capability of the network to identify and locate that terminal either in the same or a different administrative domain.. IP protocol is suitable for static networks. In wireless networks, the nodes are in constant movement. In such environment IP protocol does not result in efficient operation and instead cause lot of overhead traffic resulting in noticeable latency and unacceptable user experience. Mobile IP (MIP) refers to protocol enhancements to standard IP protocol that allows transparent routing of IP datagrams to the mobile nodes on the internet (Perkins, 2002). Some authors have defined mobile IP as a routing protocol in itself. Authors Raab et al (2005) define mobile IP as dynamic routing protocol where end devices signal their own routing updates and dynamic tunnels eliminate the need for host route propagation. This means that instead of routing tables updating in response to the terminal movement, the tunneling feature hides the address change in the routing table. A handoff in the wireless mesh network occurs when a mobile terminal 4

5 changes it point of attachment to the wireless backbone network. A change of Access Point (AP) while maintaining the connectivity is typically called a Handoff (Valko, 1999). Handoff latency measures the delay between the point in time from when MT was connected to wireless mesh backbone from a point of attachment and when it moves and is able to connect to the backbone from another point of attachment. The delay is especially significant in multi-hop transmission characteristics of WMNs due to multiple route discoveries, signaling message propagation when multiple hops are involved (Xie & Wang, 2008). Handoff Management is the process by which a mobile terminal keeps it connection active when it moves from one access point to another (Akyildiz, Xie, & Mohanty, 2004). The handoff process in the WMN can happen between two mobile nodes belonging to the same network domain, in which case the handoff is referred to as micro-mobility, or between WMN nodes belonging to different network domains where it is called Macro-mobility. Mobility Management enables telecommunication networks to locate roaming terminals for call delivery and to maintain communication as the terminal is moving to a new service area (Akyildiz et al., 1999). Mobility Management encompasses a set of tasks for supervising the mobile user terminal (or mobile Station, MS), in wireless network. The tasks are divided into registration and paging, admission control, power control and handoff (also called handover) (Giannattasio et al., 2009). Recently the Multi-Protocol Label Switching has been explored as an efficient tunneling protocol to be used in MIP. MPLS is a packet forwarding solution where labels are assigned to packets. Routing is enabled by looking at labels. Label lookup forwarding enables fast end-to-end routing without need for protocol specific lookups. What this means is that packet header does not need to be parsed at each hop. If IP routing were used, IP header has to be parsed to determine the next hop. MPLS label does away with parsing 5

6 for layer-3 or layer-2 specific protocols since it adds a label to the data before the layer-3 header and after layer-2 frame. 4. Hypothesis The wireless mesh networks are typically characterized by mobile users traversing across homogeneous and heterogeneous access networks. The access network could be organized as one whole layer-2 access network or sub divided as subnets. The layer-2 access mechanisms could be different in heterogeneous access networks. Within the wireless networks, the access methods used by mobile device to attach to the access network are different (CDMA, Wi-Fi, , etc.). As the user moves from one subnet to another or even across the IP domain, maintaining the mobile user location information becomes an overhead. The handoff management process needs to deal with location and addressability challenges of mobile nodes. Since the protocol parameters can be different across each wireless access network, it is difficult to achieve a smooth transition at each network boundary. The overhead caused by need to maintain the location and addressability information for each node, results in higher latency in wireless mesh networks compared to static wired network. MPLS inherently supports QoS features and this coupled with its lower overhead in network stack lookups should result in lowering handoff latency and reducing network resource allocation. Due to the lack of adequate IEEE standards for addressing WMN mobility issues, the vendors such as Cisco and Aruba Networks have adopted proprietary mobility solutions. 6

7 5. Scope and Assumptions The scope of this paper is limited to a survey of handoff management issues inherent in the mesh networks and application of MPLS to address those concerns. Current deployments of wireless mesh networks by vendors such as Aruba Networks and Cisco is evaluated. The intended audience is service providers who are planning to provide WMN solutions in enterprise, municipalities, and emergency services. The IEEE standard s which is relevant to mesh networks is covered. The IEEE standard is out of the scope as well as the standard which deals with media independent handover between the cellular networks and the WLAN network (De La Oliva et al. 2008). This paper is concerned with mobile terminal s terminal mobility solution also referred to as nomadic mobility. The scope of the paper mainly focuses on the intra-domain mobility issues. Thus cross-protocol handoffs which span across an administrative domain are out of scope. Wireless Metropolitan Mesh Networks (WMAN) is the deployment of mesh network over metropolitan area and not in the scope of this paper. 6. Importance The past decade has seen tremendous growth in wireless LAN deployments such as in enterprises, universities, and public wireless hotspots such as airports, restaurants, etc. However providing mobility support for mobile devices which move from one wireless access structure to another often necessitates fixed wired infrastructure as a backbone network to carry and route traffic to the Internet (Nandiraju et al., 2007). This can be expensive for service providers. WMN deployment can save CAPEX (capital expenses) for service providers because mesh nodes can self-configure and acts like a router of backbone traffic without need for expensive T1 and T3 network pipes to be laid. 7

8 Signaling and traffic management are crucial aspects of any communications network. Mobility Management plays an important role in signaling and traffic analysis as efficient traffic management load analysis studies involve studying the mobility management (Markoulidakis, Lyberopoulos, Tsirkas, & Sykas, 1997). 7. Contestable Even though there are several research papers which focus on the individual aspects of the handoff management solutions, such as solutions at layer-2 MAC, network layer, micromobility and macro-mobility solutions, a comprehensive survey paper on the handoff and mobility challenges of WMN, which evaluates MPLS technology for solving mobility issues is lacking. This paper brings together and provides a comprehensive survey of the handoff management challenges enabling a better definition of the problem areas in WMN deployment. It also explores the applicability of MPLS solution to micro mobility solution. The paper contributes to the WMN deployment scenarios by analyzing the proprietary mobility solution adapted by the service providers, especially Cisco and Aruba Networks. Because the IEEE standardization attempts for WMN mobility are still in draft stage (De La Olivia et al. 2008), this study has importance for service providers (such as Cisco and Aruba Networks) who in absence of standards have implemented their own proprietary solution to mobility problem in WMN. The IETF and IEEE work on mobility management solution is discussed in the Standards section later in this paper. Even though the handoff process in wireless networks has been studied, but real-world solution that is acceptable to service providers has been difficult to achieve (Kretschmer & Ghinea, 2010). 8

9 8. Methodology This paper identifies the micro-mobility requirements for mobility management in WMN networks, identifies the current micro-mobility solutions, discusses their inherent issues, and finally evaluates whether use of MPLS technology in micro-mobility solution can offer significant advantages. In order to accomplish these goals the paper answers the following subproblems. 8.1 Sub Problem 1 The first sub-problem is to evaluate how the wireless mesh network architecture differs from the traditional Wi-Fi wireless access architecture Data collection and research Identify the issues specific to WMN mobility and the requirements of mobility management for WMN deployments. 8.2 Sub Problem 2 Second sub-problem is to identify the handoff challenges for wireless mesh deployment and to examine how the efforts by the standards committee have helped in standardizing the WMN deployment issues Data collection and research Identify the handoff mechanisms in the WMN networks. Explain the OSI layered solutions to mobility management to understand the handoff challenges at link and network 9

10 layer. Identify and evaluate the standards committee efforts at standardizing the network and data link layer mobility solutions. 8.3 Sub Problem 3 Does MPLS offer a better solution towards solving the mobility issues apparent in the wireless mesh networks in term of reduced complexity and efficient route path selection? Data collection and analysis of results Evaluate how MPLS architecture can help towards efficient handoff in WMN. Can MPLS tunnels provide reduced complexity compared to the IP-to-IP tunnels of the traditional micro-mobility solutions in the WMN network? 9. Wireless mesh network Architecture 9.1 Traditional wireless infrastructure mode: The traditional wireless access methods such as Wi-Fi networks, WLAN, Bluetooth, and cellular networks use single hop communication; for this reason they have to be connected to an access point which connects directly to a wired backbone network. In a traditional wireless network each access point connects to a wired backbone network, which could be Ethernet LAN or any other wired access methods that in turn is connected to the Internet backbone. The AP serves to provide the mobile terminals with network access. When the mobile terminal moves from one access network to other, the user has to re-establish the connection. The major disadvantage of this configuration is that coverage can be increased only by adding new APs. 10

11 Since APs require a backhaul network, providing large wireless coverage areas becomes expensive. 9.2 Wireless mesh networks: In contrast to single hop communication the wireless mesh networks use multi-hop communication. In the WMN networks each AP is wirelessly connected to mesh routers. The mesh routers form the wireless backbone for WMN networks by routing packets between mesh routers. A WMN network is composed of Mobile MT connected to AP, from which they get the network access. The APs wirelessly connect up to the mesh routers. Two or more mesh routers are managed by mobile gateway routers (MGRs) (Garroppo, Giordano, & Tavanti, 2009). The mobile gateways routers (MGRs) are the WMN s connection point to the Internet backbone. The service providers typically use a single IP domain in which all their equipment is located. This single IP domain is divided in to subnets, with each MGR being part of a distinct subnet. Instead of being connected to a wired backbone, the MRs wirelessly connects to mesh gateway routers. Each of the MGR is an internet gateway (IG) and connects to the wired infrastructure. The entry point in the wired infrastructure is via Access routers (AR). The mesh routers are interconnected to each other and are capable of multi-hop communication. The mesh routers are the first point of network access to the Access Points because they work at layer-3 and are equipped with complete IP stack (Garroppo et al. 2009). Just like routers and gateways form the backbone for the Internet, the MRs and MRGs form the backbone for WMN. This architecture enables mobile nodes to send multi-hop messages to distance the Foreign Nodes (FN) or the Corresponding Nodes (CNs). 11

12 Figure 1: Wireless mesh network diagram AR Internet Backbone AR AR Access Routers Domain 2 Domain 1 MGR MGR Wireless Mesh Backbone MR AP Subnet C MR MR MGR AP MR AP AP AP Subnet A AP AP Subnet B MT MT MT MT Basic Service Set and Enhanced Service Set: As shown in the below figure, a simplest configuration for WMN is one which consists of Basic Service Set. A BSS consists of an Access Point several MTs. A MT is confined to a single AP. 12

13 Figure 2: BSS and ESS movement of a MT Wireless Mesh Client Network Access Point Access Point Association BSS Association MT 1 MT 2 MT 3 MT 4 BSS SUBNET A SUBNET B ESS BSS: Basic Service Set ESS: Extended Service Set Homogeneous and Heterogeneous deployments Because the WMN deployment is expected to cover a large area compared to traditional WLAN deployment, the expectation is that there will be different access systems, service providers, backbone networks which could become part of one single mesh network. The WMN architectures can be divided in two categories, namely Heterogeneous WMN deployment and Homogeneous WMN deployment. Homogeneous deployments are limited to a single service provider s domain. The entire network is managed by a single provider as a single IP domain. Single IP domain deployments contain homogeneous access network such as Wi-Fi, cellular 13

14 network, WLAN, and Bluetooth across the mesh network with similar protocols and interfaces. As shown in the figure above, the domain 1 network has its access network divided into subnets, the MT movement is within single domain but across subnets. In contrast a heterogeneous deployment will typically span across IP domains. Each IP domain may be using different access networks, assuming that the backbone is all IP; the signaling overhead and mobility management can introduce delay and latency whenever mobile terminal moves from one IP domain to another. 10. Mobility Management: A Mobility management solution for a WMN is concerned with ensuring that as the MT moves in the mesh network, the mesh router can keep track of MT s location information as well as maintaining the user s active connections to the backhaul network. From MT s perspective the network adjusts to accommodate MT s changing location, reachability, and active TCP connections. Mobility can refer to either terminal mobility, service mobility, session mobility, or vehicular mobility. These various types of mobility solutions were briefly discussed in the section 3 titled Background. In this paper the term Mobility refers to Terminal Mobility. Mobility management is a broad term referring to various aspects of mobility of a mobile terminal in the wireless mesh network. Depending on the architecture of the WMN, the entire WMN may consist of one single IP domain consisting of separate subnets or may be composed of different IP domains, each domain consisting of several subnets. The former architecture consisting of a single IP domain is usually a homogeneous wireless access environment because service provider uses a single access method to access all of its MTs. An architecture consisting of multiple IP domains is a heterogeneous environment as different IP domain may implement 14

15 different access systems in them. One IP domain may use WLAN connection where as other domain may use a cellular network to communicate between MTs. The mobility management of MTs is needed in both the homogeneous and heterogeneous environments, but in heterogeneous environments a layer-3 handoff solutions are needed, where as for homogeneous environments both layer-2 and layer-3 are candidate solutions. Depending on the movement of the MT within or out of its IP domain, the mobility solutions can be classified as intra-domain also called Micro-Mobility and Inter-domain also called Macro-Mobility Macro and Micro Mobility of mobile Terminal: When the mobile terminal changes it Point of Attachment (PoA) within an IP domain from one subnet to another this kind of mobility is referred to as micro-mobility, also called intra-system mobility. This kind of mobility is the result of network and access systems having similar protocols and interfaces. In the above shown figure 2, the MT movement from Subnet A to Subnet B is referred to as Micro-mobility (intra-system). If the MT were to move from one domain to another, the movement is called macro-mobility (inter-system mobility) (Akyildiz et al. 2004, p. 18). Macro-mobility is a result of the network spanning service providers, and the IP domains each having a different protocol stacks and heterogeneous access networks Mobility Solution: IETF introduced mobile IP as a solution for mobile terminals which needed to remain connected to the internet when changing their point of attachment. The Mobile IP uses concept of Home Agent, Foreign Agent and registration between the agents as a way to keep track of the changes in the mobile terminal location and address. The Home agent and Foreign Agents are mesh gateway routers (MGR) as shown in the figure 1. Because MGR (HA and FA in mobile IP) are 15

16 required to keep the location data for the mobile terminals in database, whenever there is a handoff initiated by MT movement, the location change information needs to be registered and propagated to mesh gateway router. This requirement for a MT to register every time it changes the PoA even within the same MR attachment causes lot of overhead registration messages. Under mobile IP solution, even when the MT moves from one subnet to other within the same home domain, the MT is required to send the registration update to the Home Agent for that domain (Akyildiz et al. 2004, p. 19). Due to inherently frequent handoffs in the wireless mesh networks, such registrations introduce latency. Due to the latency introduced by the registration and handoffs in mobile IP, the research community has adopted several variant of mobility management solution. To reduce the latency due to mobile IP registration and overhead traffic, the WMN architecture has divided the mobility architecture in two separate problem domains. The mobility of MT when it moves within the IP domain is referred to as Micro-mobility and when the mobility spans across IP domains, the mobility is referred to as Macro-mobility Micro-Mobility: Due to registration and tunneling overheads of mobile IP solution, the mobile IP has not been used in micro-mobility domain instead the Micro-Mobility solutions have adopted modified versions of the mobile-ip. When mobile terminal moves, mobile IP requires a new tunnel to be setup. The tunnel between Home Agent and Foreign Agent is needed so that mobile terminal can get the packets destined to it when it associates with the Foreign Agent s Access Point. The delay is inherent in the round trip incurred by the mobile IP as registration request is sent to the HA and the response sent back to the FA (Campbell et al. 2002). Registration and routing of the packets via tunnel between HA and FA becomes more significant when the handoffs frequency increases. Micro-mobility is characterized by frequent handoffs due to the mobile terminals 16

17 switching frequently between APs. Various micro-mobility solutions are available such as Cellular IP, Hawaii, and Hierarchical mobile IP (Campbell et al. 2002). A common approach in the micro-mobility solutions to reduce the signaling overhead associated with MT registration is to localize the signaling overhead generated by the MT movement. For this purpose typically a mobile Agent gateway is designated closer to the MT and signaling and registration need not propagate all the way up to the Home Agent gateway (Xie & Wang, 2008, p. 37). Signaling (location info) is derived implicitly or via ways transparent to a mobile terminal. The idea being that MT should be freed from signaling overhead, after it has initiated a handoff process (Caceres & Padmanabhan, 1996, p. 59). Instead of all the handoff signaling flowing to root node which is the mesh gateway router in MIP, the micro-mobility solutions implement a sort of Hierarchical router distribution. The Hierarchical router distribution helps with localizing the registration and signaling traffic because registration updates as MT moves from one PoA to another is kept with the nearest mobile router. Micro-mobility solutions such as Cellular IP, HAWAII, and HMIP tend to keep the mobility changes visible in smaller and local area. This avoids the registration messages between HA and FA from long trips, for example when HA and FA are far apart and have to cross several network devices to reach one another. Cellular IP, HAWAII, HMIP are essentially effect proprietary control messages for location management and routing within a regional area of network (Yokota et al. 2002, p. 132) Routing based and Tunnel based Micro-mobility: The Micro-mobility solutions are either tunnel based or routing based (Chiussi et al. 2002). The Cellular IP and Hawaii protocols fall under routing based solution where as Hierarchical MIP falls under tunnel based approach. Routing based approaches leverage on the IP forwarding and lookups to send the packets destined for mobile terminal to mobile terminal even when the 17

18 mobile terminal has changed its Point of Attachment (PoA). The tunnel based micro-mobile solution use the tunnels similar to mobile IP, but the mesh routers are divided into hierarchical router domain. The division of hierarchical routers helps to localize the tunnels and registration overheads (Chiussi et al. 2002) and thus avoid the mesh gateway router updates every time a handoff occurs. Whether the handoff is tunnel based or routing based, the Micro-mobility solutions need location and handoff management techniques to enable efficient handoff. Every mobility management solution has two parts, namely location management and handoff management (Akyildiz, Xie, & Mohanty, 2004). The location management and handoff management will be discussed in later sections Routing Based Handoffs Cellular IP: Cellular IP overcomes the mobile IP limitation of having to propagate the MT handoff registrations all the way to mesh gateway router. Instead of designating mesh gateway router as root router (which is always part of every routing decision), the mesh router closet to the Access Point assumes that functionality. This way when handoff occurs within the MR domain, mesh gateway router is not even involved. Cellular IP uses the packets transmitted by MT to mesh gateway router to determine the path information, thus reducing on the signaling required to keep the location database updated. Instead of a Home Agent maintaining the location database in a centralized manner, the Cellular IP requires that each mesh router keep a node to IP address of the MT. Thus each MR knows which port to forward the packet to. This hop-by-hop routing means that no single point of failure exists (Valko, 1999, p. 55). However to keep the node-to-mt ipaddress mapping updated, a periodic beacon transmission is required to be send by the MT. Since the MT may move between the APs, the mapping can become outdated. For this reason cellular IP utilizes timers to reduce packet loss due to packets delivered 18

19 to old AP. Search and lookup times are improved by using caches. Cellular IP uses IP addresses to identify the mobile terminals (Campbell, & Gomez 2000, p. 48). Figure 3: Cellular IP architecture Internet Access Router AR AR Mesh Gateway/ Domain root Router Mesh Router MR1 MGR MR2 MR3 Host Specific routing Table MT1 -- MR2 MT2 MR3 Host Specific Routing MT1 Figure : Cellular IP Architecture MT2 Home Domain HAWAII: Like Cellular IP, Handoff Aware Wireless Access Internet Infrastructure (HAWAII) allows the mobile terminal to retain its IP address as it moves in the access domain from one mesh router to other. Thus the Home Agent (usually the mesh gateway router) is unaware of the MT movement, avoiding expensive handoffs between HA and FA for each address change of the MT (Ramjee et al. 2002). Instead of extracting signaling information such 19

20 as node reachability and port to next-hop address mapping from MT s normal data traffic the Hawaii protocol requires the mobile terminal to send explicit signaling messages so that HA can use these signaling messages to determine the IP routing path to MT. Location information (i.e. mobile-specific routing entries) is created, updated, and modified by explicit messages sent by mobile Host (Campbell et al. 2002, p. 73). HAWAII works on hierarchical network segregation by diving access network in domains. Each MT has a home domain and foreign domain. Inside a Home Domain, a domain root router is designated. Each MT send periodic infrequent signaling message to root domain router. This establishes a path setup between Domain root router and MT (Ramjee et al. 2002). By selecting only a few designated routers which participate in the path setup updated messages, the signaling traffic overhead is minimized. Further HAWAII is similar to Cellular IP in respect that the MT retains its IP address as it moves within the home domain or within its foreign domain. But rather than using hop-by-hop tables like in Cellular IP where each node maintains next-hop table with Host IP address and port mapping, the HAWAII location management uses IP address forwarding technique, so as far as routing is concerned HAWAII uses IP routing mechanism to reach mobile host Tunnel Based Handoffs Hierarchical MIP: Hierarchical micro-mobility solutions do not use IP addresses or hop-to-node mapping to reach from gateway to mobile terminal, instead they employ tree like structure of Agents, and each Agent maintains destination MT s address to next Agent address mapping. Thus the location database is distributed across several Agents. Once a MT registers with it s HA, after that it maintains location database update only with its immediate next Agent (Campbell, Gomez 2000). This way location database updates when MT moves is visible only to the lowest Agent in the hierarchy of the Agent tree. 20

21 Location Management: Reaching the roaming MT for delivering the packets destined to is a challenge that location management tries to solve. In traditional wired infrastructure when an IP device moves from one subnet to other the TCP connections are broken. TCP connections work on static location addressing; IP addresses are assigned in a hierarchical manner. So when an IP device in a wired network moves from one subnet to other, its new subnet will assign it new an IP address. With the new IP address, the TCP connections need to be re-established. Whereas when a MT moves from one subnet to another, the expectation is that the MT will still be reachable with the same connection end points as if it has not moved from its original location from a network perspective. Location Management involves keeping track of the MT while it moves from one AP to another. This is necessary because packets destined for MT need to be routed to its current location and not its home location. To make it possible to route MT A s packets to a location that is MT A s current location, there is a need for maintaining a database with MT s current foreign and home location information. The database needs to maintain the mapping and every routing decision will necessitate a lookup for current location before packets destined to a MT can be routed. Each of the three Micro-mobility solutions maintains the location database at different mobile nodes in the mesh network. Also the composition of the location database varies depending on whether mobility solution. In case of Cellular IP the location database contains the mapping of destination MT IP address and interface port use to forward the packet. In case of HAWAII, the location database consists of IP address to next mobile node address. In case of Hierarchical MIP, specific set of agent nodes maintain database of destination mobile node address and next Agents address for forwarding the packet. Location management involves keeping the location data refreshed and current. This necessitates paging and beacon messages 21

22 between mobile gateway nodes and MT (Mohammad et al. 2009, p. 679). The location management can be classified into (Xie & Wang, 2008): Address Management: roaming. Address management involves keeping track of the identity of the MT while it is Movement detection: A MT has to know that it has entered a new location or area. This can involve either an active periodic probe request message by the MT to the mesh routers or it can be a passive router advertisement beacon message from mesh routers to the MTs. When this happens the MT knows that it has entered a new area and thus registration has to begun. MTs can look at the subnet mask to identify if it has moved under new AP and new MR or just done a layer-2 handoff Paging: Paging involves determining the MR of the MT in question. This is needed because the MT gets its network connection from its associated MR Handoff Management: Handoff Management deals with keeping the connections active as the MT changes its Point of Attachment. Handoff of the MT involves update of the location database because MT s new location needs to be communicated to the mesh routers and mesh routers contain the mapping data to reach the MT. The various solutions to handoff depend on which OSI stack layer the handoff is being done. The handoff solutions fall in three categories and they are link layer handoff, network layer handoff and cross layer handoff. There are tradeoffs in each solution whether it is link layer, network layer or cross-layer, these handoffs which will be 22

23 discussed in subsequent sections. There are two types of MT movement in the wireless mesh network. In first case the MT may move from attachment to a MR to another MR. In this case the MGR attachment still remains the same. If the MT changes its MR attachment then it performs Link-layer handoff, where as when the MT changes its MGR attachment it performs a Network layer handoff (Xie & Wang, 2008, p. 39) Design Issues for Handoff Management: Handoff detection: A MT needs to detect a handoff is necessary before it can initiate a handoff. Forced Handoff occurs when the point of attachment of the MT or other mesh Hosts such as routers changes in the mesh network. The MT initiates procedure to find new MR attachment. Unforced Handoff occurs when MT finds a better path to the MGR and thus initiates a handoff to connect to new MR which may or may not lead to new MGR attachment (Xie & Wang, 2008, p. 39) Mesh gateway router selection: If a handoff occurs and MT finds a new MGR point of attachment then a network-layer handoff needs to be initiated by the MT. This leads to the third handoff issue namely, the QoS maintenance QoS Maintenance: If during the handoff a MT finds several MGRs to choose from then the MT needs to consider the QoS maintenance issues. The MT needs to make sure that the minimum QoS that it had with previous MGR is what it can get with the new MGR. The QoS guarantee involves maintaining resource reservation along the path that the MT will reach the new MGR. 23

24 The MT s change in attachment to the mesh Network can result in link-layer or networklayer handoff. The handoff can be broadly classified as Link-Layer handoff where the MT changes its association or PoA from one mesh router to another and the Network-Layer handoff where the MT changes its PoA with respect to the mesh gateway router, also called the Access router. The MT gets its network layer address from the MGR/AR Link Layer Handoff and Network Layer Handoff Issues: Link Layer Handoff: The link layer lives in second layer of the OSI protocol stack. This layer is responsible for node to node movement and the addressing is based on MAC addresses instead of IP addresses. Link layer handoff happens when a MT moves away from the radio range of one AP to another. The MT and its attached AP form a Basic Service Set (BSS). At this point it leaves one BSS and enters other BSS. During this link layer handoff the MT exchanges management frames with the new AP to form a new BSS. Forming new association, exchanging credentials leads to latency in handoff. When MT moves in the subnet such that its connection to the serving MGR remains the same, the handoff is still link layer handoff. This is also known as access handoff or intra-system handoff because the devices attached to MGR, such as MRs, are link layer devices connecting to the MT by link interface. The MT may move from on one AP to other AP which maps to a different MR. This is still a link layer handoff, as long as the serving MGR remain the same (Cisco, 2009). The roaming in the layer-2 network can be of two types: 24

25 Layer-2 roaming in layer-2 network: When a mobile terminal moves from one AP to another and both the APs are still attached to the same MGR and in the same subnet, the roaming is called layer-2 roaming in layer-2 network. This type of deployment allows the mobile client to roam from one AP to another without needing a change in client IP address (Cisco, 2009). Figure 4: Layer-2 roaming at Layer-2 OSI layer MGR MGR Wireless Mesh Network Mesh Gwy Routers MGR MR MR Mesh Routers Access Point Access Point Association BSS Association MT 1 MT 2 MT 2 MT 3 BSS S U B N E T A 25

26 Layer 3 roaming in Layer-2 network: When the mobile client moves from being attached to one AP to another AP which is controlled by a different MGR and in a different subnet, then this kind of mobility is called layer 3 roaming in layer-2 network (Cisco, 2009). This type of mobility supports mobility from one type of access network to a different type. For example Subnet A could be a WLAN radio network and Subnet B a cellular network. Thus heterogeneous access networks are supported by layer 3 roaming. Figure below illustrates the layer-3 mobility: Figure 5: Layer-3 roaming at Layer-2 OSI layer MGR Wireless Mesh Network MGR Mesh Gway Routers MGR MGR MR Access Point MR Mesh Routers Association Association MT 1 MT 2 MT 2 MT 3 Subnet A Subnet B 26

27 Link layer handoff delays: The link layer handoff follows the handoff procedure. This procedure has several steps (CodeAlias, n.d.) Scanning delay: The MT scans for suitable AP to connect to. This is needed when current AP s SNR is low or other conditions where continued attachment will result in signal loss Association delay: After selecting suitable AP, the MT has to associate with the new AP. The association, also, allows the new AP to inform the link layer devices (bridges, switches) to update their L2 table so that packets in destination to the STA get forwarded to the new location Authentication delay: The new association between MT and AP need to be authenticated X enabled APs only accept 802.1X frames from non-authenticated STAs. The 802.1X frames contain EAP messages that are forwarded by the AP to a back-end RADIUS server over the RADIUS protocol. More detailed description of the delays due to authentication can be found in CodeAlias (n.d.). A link layer assisted handoff has been proposed by Yokata et al. (2002). This architecture requires a layer-2 bridge to filter the MAC addresses before forwarding to the home domain. When a MT finds a neighboring AP with which it associates, the neighboring AP broadcasts the MT s MAC address in its local segment. This causes the MT s MAC address to be registered in the MAC bridge (the MAC bridge filters the packets before handing over to the Home/Foreign domain root router). With a 2 port bridge, the subsequent packets will now flow from HA to MAC bridge and to Foreign Agent to new AP and thus to the MT. After the MAC bridge port 27

28 mapping timer expires, the flow is tunneled from HA to FA and to the new AP and the MT. With the link layer approach using the MAC bridge, the need for HA registration does not need to happen before the MT in new domain starts receiving the packets. The registration delays are thus not a requirement before MT can start communicating after handoff. This avoids the registration related delays evident in the mobile IP before communication can resume between the MT and the Internet Network Layer Handoff: Mobile IP Mobile IP (RFC 3220) which is IETF s specification for network level mobility, incurs lot of registration overheads, suffers from triangular routing problem, packet loss during handoff, and does not support paging (used to locate the terminal s mesh router). In contrast to the link layer handoff, the network layer handoff allows handoffs to occur between two domains or subnets. When a MT moves from one subnet in a domain to another, often the Access router needs to update its routing entries. This scope of the handoff is now not within a single domain but across domains. To maintain the MT s IP continuity in such cases needs MT to execute a network layer handoff. PMIPv6 (Proxy mobile IPv6) is one network layer handoff protocol which a proxy agent helps the MT execute network layer handoff Proxy MIPv6 Proxy MIPv6 (RFC 5213) (Gundavelli, et al, 2008), the IETF proposed standard, is a network-based local mobility protocol. It defines Mobile Access Gateway (MAG) and Local Mobility Anchor (LMA). MAG is the access router which provides access to the Internet, where as LMA is local home agent router (Lee & Min, 2009). The LMA provides the MT with its 28

29 anchor point connection to the network layer of the mesh network. The MAG assists the MT in the handoff by detecting the movement of the MT from one subnet to another. When this happens the MAG searches for the LMA in the new subnet and finds if the new LMA has the binding Cache entry for the MT. The binding cache entry is mapping between the MT s Home address and its Care of Address in the new subnet (RFC 5213, 2008). If not then MAG sends the binding update to the new LMA in the moved to subnet. A tunnel is established between the MAG in the home domain and the LMA in the visited domain. This enables the MT to be addressed with its Home Address even though it has moved to a different subnet and has separate Care of Address, called the Proxy CoA (RFC 5213, 2008), from the visited subnet. The predecessor to the PMIPv6 namely the MIP required stack modification in the MT because it needed the MT to be part of the handoffs. However the proxy MIPv6 which is network based handoff takes that responsibility away from the MT and network layer instead handles the handoff for the MT (Garroppo et al. 2009). There are many mobile devices which are not necessarily mobile aware, so a implementation was needed which could free up the mobile terminal from handoff mechanism. Thus the proxy MIP was born. In the PMIPv6 implementation, the entire handoff is performed for the MT by the network layer and initiated by the MAG. Network layer involvement has known to cause handover latency and packet loss before the MT in the new subnet can start receiving the packets destined to it (Lee & Min, 2009 p. 1085), for this reason modified version of PMIPv6 namely PFMIPv6 (Proxy Fast mobile IPv6) was developed. In PFMIPv6, the handover procedure is done before the MT executes a handoff. This saves on latency and packet loss is avoided. In the PMIPv6, the MAG detects the MT movement and then initiates the network handoff by establishing tunnel with foreign LMA. In PFMIPv6, the MAG depends upon the link layer trigger to initiate the tunnel before handover is 29

30 started. For example the MT link layer can detect the presence of a nearby AP which has stronger signal and trigger signal to its home domain MAG that a handoff is necessary. The home domain MAG sends MT identifiers such as MT id, MT-LMA id etc to the foreign LMA in Handshake Initiate message. Thus the Foreign LMA has handover parameters that will be needed for binding before the handover is executed (bi-directional tunnel between MAG and foreign LMA establishment) (Lee & Min, 2009) Mobility requirements for handoff: When mobile clients move across the network, the following need to be incorporated for a successful handoff: Location Database: Location database is information about the mobile terminals association with AP and MAC address. The database is located in the MGR s client database. The entries stored for each mobile terminal are client MAC and IP address, security context and associations, QoS, WLAN and associated AP (Cisco, 2009, pp2-17). As the mobile terminal association with AP changes, the MGR s client database is updated. This update is either the responsibility of MGR and happens at the network level by communication between the MGR s or can be the responsibility of the mobile terminal to update the client database Move Discovery: There are two ways for move discovery. The mobile client s layer-2 service detects disconnect at layer-2 level and issues a notification called Media Sense to the Windows OS (Cisco, 2009, pp 12-4). This enables the mobile terminal to recognize a move and request new IP from DHCP server. The other way is for mobile Client to receive a Foreign Agent advertisement. 30

31 Within a subnet the mesh router periodically sends a beacon message to all the devices advertising its existence. This way any visiting mobile client can detect a change in association with AP in the visiting subnet, request new IP and initiate a handoff Location discovery: Once the move discovery is done and if it is a move to a different subnet, then either a tunneling or routing update method is used for routing packets to the mobile client in the new subnet. For tunneling method the mobile Client has to update the Host MGR about the MTs new AP association and new Foreign MGR s address. The MAC being a flat assignment remains same Challenges of Link Layer Handoff: As we note above during the link layer handoff involves scanning, management frame exchange, and authentication delays. There could be several APs served by a single mesh router, all the network layer-2 elements and when the AP is involved in a handoff, the signaling messages have to pass through all the layer-2 devices generating lot of network traffic and overhead. Thus pure layer-2 handoff is not an efficient solution. Combination of layer-2 triggers with layer-3 handoff establishment is a good working solution for handover. 31

32 11. Standardization efforts in wireless mesh network: 11.1 Current mobility standards and their status IETF mobility standards The Mobile IP (MIP), which is the RFC 3344 (Perkins, 2002), is the current IETF standard for supporting mobility on the Internet. IETF introduced mobile IP as a solution for mobile terminals which needed to remain connected to the internet when changing their point of attachment. But the IETF Mobile IPv4 suffers from longer handover delays mainly due to AAA (authentication, authorization and Accounting) signaling, IP address configuration, and packet loss during handoff (Adibi, Naserian, & Erfani, 2005). To alleviate the Mobile IP related issues, a hierarchical Mobile IP has been proposed by IETF as proposed standard in October 2008 called the Hierarchical Mobile IPv6 Mobility Management (HMIPv6) in RFC Successor to the Mobile IP was the Proxy MIPv6, which is the IETF proposed standard (RFC 5213). The PMIPv6 handles all the handoff related signaling in the network layer nodes, this way mobile terminal does not have to get involved in the signaling and thus overhead is reduced. There are many mobile devices which are not necessarily mobile aware, so a implementation was needed which could free up the mobile terminal from handoff mechanism. Thus the proxy MIP was born. The Proxy MIPv6 is discussed further in the section Mobile IP with Regional Registration (MIP-RR) was proposed in RFC 4857 to reduce the registration delays. MIP-RR uses hierarchical levels for home and foreign agent, so that registration related signaling can be handled locally and does not need to travel across to one single common home/foreign domain agent for the entire domain. This reduces the signaling traffic at the upper level home agent. 32

33 IEEE Standards for mobility management The IEEE standards committee established the Task Group named s TG for the s amendment to the existing WLAN standards to address the Wireless Local Area Mesh Networking. The s is presently in the draft stage. The latest activity in the s task group has been to forward the draft s document to the RevCom (Review Committee) (IEEE, 2011). The working group is IEEE s effort to standardize the mesh architecture in Wide Metropolitan area Networks (WMAN). The 3G cellular network s 3GPP program (Third Generation Partnership Program) uses IETF defined Mobile IP for its mobility solution in cellular network). The 3GPP networks handle the issue of mobility in two ways; the 3GPP2 uses Mobile IP for IP mobility (i.e., terminal mobility) management and SIP for session mobility management (Munasinghe & Jamalipour, 2008) IEEE s standards Overview A device which confirms to the MAC and PHY specifications is identified as a Station. A Station which can acts as a central device for other wireless LAN stations is called an Access Point (AP) and this topology is called a Basic Service Set (BSS) (Hiertz et al. 2007). A BSS forms a single hop network and all the devices on the basic service set depend on the central AP to relay frames to each other. When several APs interconnect with each other, they form an Extended Service Set (ESS). Stations can roam between APs in an ESS area. The IEEE s takes this topology further and defines a mesh network. It defines a mesh Point (MP) as a mesh network element which can relay frames between MPs using multi-hop communication. The s link management protocol is used to discover peer MPs which can then establish the link 33

34 layer connection with each others. The peer MP discover can involve passive scanning, which is based on listening to the beacon frames from nearby MPs or active scanning by sending probe request to nearby MPs. Once the neighborhood MPs are identified, they can form a mesh network. The IEEE s amendment s describes the necessary functions to form a wireless mesh network. The s is an MAC layer approach to multi-hop communication between the mesh Points, versus the earlier approaches to using network layer for multi-hop frame exchanges (Carrano et al. 2011, p. 53). Typically s defines a wireless mesh network with point of view of mesh routers being fixed in their location relative to each other. This contrasts with mobile Ad hoc Networks (MANETS) in which there are no fixed routers. In a MANET the routers are mobile and ad hoc without the need for infrastructure of mesh routers for routing frames. The mesh routers in the WMN are responsible for exchanging the frames. Since the mesh routers form the backbone wireless network, they need to constantly update their routing tables. There are two route discovery methods for updating the routing table (Carrano et al. 2011, p. 54): Proactive Routing Updates: In proactive approach of the route discovery, the mesh routers are constantly exchanging the routing information regardless whether there is a need for data transmission between the mobile terminals. The proactive approach tries to keep the routing tables updated. As evident this approach results in the excessive traffic, and since the effort is to keep the routing tables updated at all the time and since the nodes in wireless network are mobile, the tables can quickly get outdated and thus greater need to exchanges routing entries frequently resulting in excessive traffic overhead. Two examples of proactive routing are Optimized Link State Routing (OLSR) and Destination-Sequenced Distant-Vector Routing (DSDV). 34

35 Reactive Routing Updates: Reactive approach to route discovery takes an on-demand approach. When a mesh router finds that there is a need for the routing tables to be refreshed, it initiates the path discovery mechanism. Examples of reactive routing protocols are Dynamic Source Routing (DSR) and Ad hoc On-Demand Distance Vector (AODV). Recently however the focus of the routing protocols for WMN has shifted to layer-2 multi-hop routing. Layer-2 allows easy embedding in the network cards for the mobile devices (Carrano et al. 2011). The s defines three entities that make up a wireless mesh network, the mobile Station (MT), the mesh Station/router (MR), the mesh Access Point (AP), and the mesh gateway router (MGR). The mesh gateway router is the portal for the mesh network to connect to other networks. Each element in the mesh network needs an identity and this identity is called the mesh ID. Mesh ID along with the path selection protocol in use in the mesh network and the path selection metric constitutes the unique mesh Profile of each mesh element. All the elements in the mesh cloud (formed by interconnected mesh routers) share the same mesh Profile. mesh beacon frames are used by the MR to discover peer MRs. MRs form mesh peer links with each other, each link identified by the MAC addresses of participating MRs and link identifier Adaptive wireless routing Recently there have been number of Adaptive routing protocols that have been suggested fir wireless mesh networks. Traditional wireless network protocols such as OLSR and AODV use static parameters which are preset and may not be suitable for all network conditions, resulting in network degradation performance when used in environments that they were not 35

36 designed for (Azzuhri et al. 2010, p.145). The adaptive WMN routing protocols will dynamically change certain parameters that are initially pre-set in the network. Adaptive AODV defines three mobility levels low, normal and high. By checking the number of 1-hop neighbors it will use appropriate mobility level (Azzuhri et al. 2010, p.145). So if it finds that there are a lot of 1-hop neighbors with which it has association it will reduce the HELLO message interval. Similarly in adaptive routing version of the OLSR routing (OLSR is link state routing). Based on the link condition the periodic link messages are either reduced or increased WMN Challenges for the Standards and Research bodies: With the s amendment in the specifications, IEEE redefined the MAC layer so that multi-hop communication is enabled in the MAC layer for the mesh routers participating in the mesh networks (Carrano et al. 2011). The challenges for WMN are: Routing protocol: Due to inherent stability issues of the wireless links, the wired routing protocols are not suited for wireless mesh networks. For this reason ad hoc wireless routing protocols are used in the WMN. These protocols broadcast frequent routing messages to gather routing metrics in order to provide better QoS on the mesh (Hiertz et al. 2008). In wired links the layer-3 routing protocols can access the neighboring links and determine the routing metrics, enabling them to determine the routing paths and update the routing tables, but in the wireless mesh environments, due to presence of radio links which IP layer cannot directly access, using IP protocols in wireless mesh networks in a challenge. For this reason and since the MAC layer is closest to the radio access physical layer, the standards organization foresee the integration of the routing and the frame forwarding in to the MAC layer (Heirtz et al. 2008). Thus L2 routing which supports 36

37 MAC address based mesh path selection and forwarding using radio aware routing metrics is what is proposed as part of the s standard. An essential component of the routing solution is the use of metrics to determine the preferred route between source and destination (Faccin et al. 2006). The airtime metrics used for radio links in the wireless mesh networks should ideally be multi-dimensional, taking into account the link condition, bandwidth, QoS considerations, power requirements etc Thus the WMN networks call for routing protocols operating at layer- 2 MAC layer Link Management Mesh points use passive scanning or active beacon transmission for finding the candidate peer mesh point. Once the candidate mesh point is identified and mesh link established, the airtime metrics is calculated (Heirtz et al. 2007) Path selection The airtime metrics of the peer mesh links calculated above is used during the path selection. Path selection protocols at layer-2 are mentioned in the section. The path selection helps mesh point determine the path to the root or portal mesh point QoS: In case of mesh Networks two main QoS issues are identified, namely for access network traffic and backbone traffic (Faccin et al. 2006). For both of these traffic networks, since the routing solutions for the WMN operate at layer-2 and layer-2 MAC layer is responsible for forwarding frames from hop to hop, the challenge is to provide consistent QoS over end to end link which covers multi-hop route path. Cross layer routing solutions have been proposed to help layer-3 routing take advantage of the layer-2 proximity to the radio medium for providing QoS 37

38 over multi-hop links and at the same time improve up on the route optimization that comes with layer-2 triggers. 12. MPLS for Mobility Management: 12.1 MPLS for routing: In the conventional IP based routing, the packets are assigned to the stream based on the longest prefix match of the destination IP address. This longest prefix match is done at every hop at intermediate routers. At every router the network header is required to be parsed to compute the longest prefix match. In the inter-domain routing where the routing tables are much bigger than the inter-domain routing, this lookup and computation for longest match can add substantial overhead to routing process. Multi-protocol label switching (MPLS) uses labels attached to the packets for routing from hop to hop, the packets are assigned to stream based on the packet label, unlike the IP address based routing where at each hop the packet may be assigned to a different stream based on the longest prefix match computation. MPLS allows use of traffic engineering and policy based routing. Because the routing is based on labels and not the destination address, policy routing defines the stream to which the packets can be attached. The routers which assign the labels to the packets are called the Label Edge routers (LER) and the routers which use the labels for forwarding the packets are called the Label Switching routers (LSR). The path between two LSR is called Label Switched Path (LSP). Packet with labels mapped to a common Forward Equivalence class (FEC), will all flow through the same route. 38

39 12.2 Mobile nodes name and location identity with IP routing: In the IP world, a node is identified and named by the IP addressed assigned to it. The IP address identifies not only the identity but also the location binding of a network node. This works fine in static wired networks where the routers are stationary. In wireless networks where the mobile nodes are mobile, this kind of routing which is based only on IP address does not work because as the mobile nodes move for example from a point of attachment in subnet A to point to attachment in subnet B, the change in its location in IP domain necessitates an IP address changes. The tying of the IP address with location information is needed in IP world because IP addressing is hierarchical. Without the IP address change the other hosts on IP network would not be able to communicate with mobile node. The IP address change of the mobile node causes all the routing tables in the upstream network to change and this result in delays when handoff needs to take place. The mobile IP tries to solve this problem by keeping the moving mobile node s address constant as it moves from one domain to other. This is achieved by using mapping between its home and foreign addresses at the network anchor point. However bi-directional tunnels are necessary to be setup between the mesh gateway router and the home anchor point. This tunneling introduces delays because frequent mobile node movement causes registration messages to be sent to mesh router gateway How MPLS fits in the mobility solution: MPLS as routing layer binding the layer3 and layer2: Mobility solutions have been explored which try to separate mobile hosts IP identity from the mobile nodes location identity. One such solution by Sethom, et al. (2004) implements 39

40 a virtual layer between the layer-2 and layer-3 to separate the mobile nodes physical internet identity from its location identity. This virtual layer uses MPLS routing mechanism for independence for layer-2 and layer-3 routing. The MPLS helps achieve independence between the layer-2 forwarding mechanism from the application identity (IP address) (Sethom et al. 2004). As noted earlier in the paper the layer-3 network handoff introduces latency because the routing updates need to be made in the routing table for each handoff made by the mobile terminal. The layer-2 handoff can reduce this handoff latency, but suffers from broadcast messages and probe messages in entire layer-2 domain. The MPLS routing mechanism lies in between the layer-3 and layer-2 routing solutions, on one hand it provides routing by using fast label look-ups at each hop without needing label decoding function (unlike network headers that are decoded at each Layer-3 devices), thus also providing layer-2 like fast MAC lookup functionality Static binding between the mobile nodes identity and physical location: MPLS breaks the static binding between the mobile nodes identity and its physical location in the internet addressing scheme. When a node physically moves from one location in the internet addressing scheme to another, its IP address also has to change. MPLS breaks this static link between the node identity and node location by using labels to do the routing function while maintaining fixed IP address as its identity. The dynamic binding between labels and terminal s physical points of attachment breaks the static association between the location and identification in the current internet architecture, and enables transparent and efficient mobility. Routing is then accomplished by labels instead of IP (Sethom et al. 2004, p. 66). The static binding between the nodes identity and at the same time its use for routing packets destined for 40

41 the node creates limited scope for using the routing entry for anything other than routing function. By decoupling the routing entity from the nodes identity, the MPLS labels can be used for multiple functions at the same time. For example labels can identify the type of service that packets should receive, the resources that the packet stream in the FEC should be allowed to use, and the routing policies based on labels MPLS as hierarchical isolation between mobility anchor point and the mesh routers: As noted in the section micro-mobility, the micro-mobility solutions such as Hierarchical MIP, Cellular IP, and HAWAII help in reducing the handoff delays by partitioning the mobility domain into local mobility regions. This way the mobile terminal movement is localized and home agent router is not involved for every handoff thus the registration delays caused by the registration between the mobile terminal and the home agent router is avoided. MPLS solutions for micro-mobility take similar approach in localizing the mobile agent movement updates. The gateway home agent which is the anchor point for mobile terminals and is the designated labeled switched router, the mesh router serving as an acting base station and mobile terminal are the only nodes involved in the registration and handoff process MPLS with traffic engineering support in wireless mesh networks: MPLS brings with it the traffic engineering concepts which are critical to ensuring a user acceptable QoS experience when moving from mobile terminal s home domain to foreign domain. With WMN networks requiring carrying voice, data, and video there is a need for reliable QoS mechanism which will not add signaling overheads to what can be sometimes be 41

42 unstable and bandwidth starved radio links between the mobile nodes in the mesh network. The MPLS RSVP-TE can be used to implement an IntServ model. In the case of RSVP-TE the end host-to-host connections are replaced by reservations between network elements. This reduces the signaling overhead. By employing signaling protocols such as RSVP-TE a LSP can be setup and managed dynamically by using dynamic or static routes. A LSP can be configured to provide QoS guarantees and to follow automatic reconfiguration when a failure appears of network state changes (Boringer et al. 2005) MPLS tunnels in the Micro-Mobility solution: As discussed in the section under Network layer handoff, when the mobile terminal moves from one subnet to another, the first step that MT does it register with its Home agent. The registration involves making the Home Agent aware of the MT s care of address which could be the Co-located COA or Foreign Agent s CoA (RFC 3344). From that point on when HA receives packets destined for the MT, it establishes tunnel between itself and the visited Foreign agent. Even though the packets destined for MT s home IP address is encapsulated in the IP header with CoA, the routers that are part of the tunnel have to re-examine the network header for determining the next hop using longest prefix match. If this tunnel is replaced with the Label Switch Path, then the label lookup will need to be done only at HA and at FA and not at the routers in between these two edge routers. The HA and FA act as Label Edge routers (LER) attaching and stripping the labels respectively. Once the label is stripped the conventional IP address routing is used to reach the MT in the visited network. This saves 16 bytes per packet transmitted (Boringer et al. 2005). Thus the MPLS is an efficient light-weight tunneling technology, using Label switched paths between the home and foreign agents an overlay network is efficiently created and managed (Chiussi, Khotimsky, & Krishnan, 2002). 42

43 MPLS by providing LSP paths create an overlay network, this means that existing network nodes by incorporating label switching capability turns the internet network into a more efficient label based switching network riding on top of the existing IP network MPLS based handoff: The handoff mechanism in MPLS based micro-mobility solution uses established LSP paths between the Home agent and the foreign agent. When the mobile terminal moves from one subnet to another, it sends the registration message to the new mesh router, in response to the MR s agent advertisement message. The new mesh router registers with the gateway home agent using registration request message. The LSP path is established between the gateway agent in the domain and the foreign agent in the moved to subnet. The gateway updates its label database and sends registration reply message to the new MR. The new MR sends the registration message to old MR as well, so that old MR can send the mobile terminals packets to new MR MPLS compared to traditional micro-mobility solution for registration and handoff delays: Micro-mobility solution using MPLS defines a new mobility agent called Label Edge Mobility Agent (LEMA). LEMA functions as any other Label Switched router (LER) in that it maps the IP address of the mobile host/agent to a FEC. The FEC itself consists of next-hop and the MPLS label. Thus FEC defines the LSP. In addition to a regular LER functionality, the LEMA also accepts registration message and on receiving one, it maps the IP address in the registration message to a FEC class. Thus the LEMA can on fly define new LSP when it receives registration request message from the mobile host. LEMA is a mobility enhanced version of LSR (Chiussi, Khotimsky, & Krishnan, 2002). The LEMA is any LSR which has the mobility 43

44 function added to it. When a mobile node moves from one AP to another in the LEMA domain, on receiving the registration message from mobile node, the LEMA enabled node maps the new CoA of the mobile node in the new subnet to a new FEC, thus defining new LSP path to the mobile node dynamically. The overhead to create new LSP is much less than overhead of creating routing table entry in the traditional micro-mobility solutions like HAWAII, and Cellular IP LEMA network and attachments As the mobile node in the MPLD network moves from one AP to another, the LSP paths are setup dynamically based on which advertised LEMA path the mobile node chooses. In contrast with other traditional micro-mobility solutions (such as Cellular IP, HAWAII) the MPLS based solution allows the mobile terminal to chose a predetermined LSP path based on which LEMA it chooses to register with. In the figure below the registration and path selection is described (Chiussi, Khotimsky, & Krishnan, 2002). 44

45 Figure 6: LEMA registration and path selection 8 LEMA 8 4 LEMA 4 7 LEMA 7 5 LEMA 5 6 LEMA 6 LEMA AR 1 AR 2 AR 3 3 LSP path advertisement LSP Path: (1, 4, 7, 8) LSP path advertisement LSP Path: (2, 4, 7, 8) LSP path advertisement LSP Path: (3, 5, 7, 8); (3, 6, 8) Access Network 1 Access Network 2 MT MT When MT is attached to AR1, its LSP path is (1, 4, 7, and 8). When the MT moves and attaches to AR2, based on the advertisement from AR2 about the available LSP paths, the MT can chose path (2, 4, 7, 8). As can be seen the change is only at the first node of attachment from LEMA node 1 to 2. When the MT completes its link-layer attachment to AR2 and parses the advertisement (2, 4, 7, 8), it recognizes that the LEMA node match is at level 2 at node 4. Thus in this case it needs to only change attachment at node 4. Thus the MT sends a registration message to LEMA node 4, with change request to send the packets destined for MT to LSP path 45

46 (2, 4) instead of (1, 4), the rest of the LSP path (4, 7, 8) remains same for the MT whether it is attached to AR1 or AR2. When the MT moves from coverage of AR2 to AR3, the AR3 advertises the LSP path (3, 5, 7, 8) and (3, 6, 8). When MT compares these new paths to the current LSP path, it sees that change needs to happen at level 3 node. The MT has the option of choosing any one of these LSP paths. The MT sends registration messages to the nodes which changed in the handoff so that they insert the mapping of the MT s IP address to the new FEC along the LSP path (3, 6, and 8) Key Points in MPLS label routing compared to traditional micro-mobility routing: Simplified Registration: The registration process is simplified, because when MT moves from one access point to other, it just need to change the LSP path that it can chose based on path selection criteria and this involves simple label swap and new FEC generation. The traditional route updates in traditional micro-mobility solution is more time consuming, thus increased latency during handoffs Flexible Path Selection: The mobile terminal based on the advertisement received from the LEMA need to make a path selection and then swap the Label for generating new FEC at the root LEMA for Path change. In traditional approach the MT would have to propagate the mapping change all the way to Home Agent. Though the Hierarchical approaches help in localizing the registration propagation, the overhead is increased number of agents to keep track of local mobility. In 46

47 MPLS, tunnel redirection, which is a crucial ingredient of any mobility scheme, happens quickly, at a change of a node in a single node in the network (Chiussi, Khotimsky, & Krishnan, 2002) QoS Guarantee and Link Reliability: LSP paths can setup RVSP-TE QoS guarantee mechanism which is lacking in the traditional routing. Link reliability is much improved in MPLS networks. This is because when a node experiences failure in IP routing network, since in the traditional IP addressing network the packets travel nodes are all the link Packet loss during handoff: Traditional micro-mobility solutions experience packet loss when MT moves from one AP to other. When the MT is in movement the packets that are received by the old AP may be lost because it is not able to contact the MT. Some implementations of traditional micro-mobility solutions use buffering at all of the nodes which avoids this problem. MPLS based nodes avoid this problem due to implementation of the redirect message to the old node to send the packets to new LEMA registered node to which the MT has moved to IP-to-IP Tunnel overhead is avoided: The IP-to-IP tunnels are completely avoided with MPLS implementation. Instead of tunnel establishment during handoff, the nodes setup LSP path and MT has ability to chose which LSP path, instead of hierarchical IP address path setup in traditional micro-mobility solutions based on network hierarchy. 47

48 12.7 MPLS throughput compared to other micro-mobility during handoff process Xie et al. (2003) have studied the throughput that the MPLS based routing can achieve when compared to the other micro-mobility solutions such as HAWAII and Cellular IP. In their measurements, the MPLS based micro-mobility can support throughputs of mbps compared to throughput of 1.22 mbps when the mobile nodes speed is 15 meters per sec during the handoff process. At higher speeds such as 40 meters per sec. the throughput for MPLS based WMN falls gradually to mbps where as for HAWAII based WMN the throughput falls significantly to 0.8 mbps. Figure 8: Throughput in mbps vs. the mobility handoff speed using MPLS and non-mpls micro-mobility solutions. MPLS vs. Traditional Micro-mobility Handoff Performance HAWAII (mbps) MPLS (mbps) mps 20 mps 30 mps 40 mps 48

49 13. Vendor implementation for mobility Solution There have been several wireless mesh implementations in the commercial industry. Aruba, Cisco, and other have implemented their own proprietary solutions for mesh networks Aruba Network s wireless mesh network solution Aruba Network s proprietary wireless mesh network implementation is called the Airmesh multi-service mesh solution. It uses Adaptive wireless Routing (AWR) and MobileMatrix in its mesh network deployment Adaptive Wireless Routing (AWR): AWR is Aruba s proprietary routing solution for wireless networks. It works with layer-2 triggers for RF awareness and layer-3 for network routing intelligence to provide routing solution that is optimized for wireless network. A routing protocol which is entirely layer-3 based can overwhelm the network, because it needs routing updates to be sent to all the layer-3 modes. In fast moving mesh network where nodes are highly mobile such routing update overheads introduces significant latency. Aruba s AWR routing uses cross layer approach. The layer-3 intelligence is combined with layer-2 information such as radio link strength and airtime matrices discussed earlier in the paper to calculate. Such layer-2 parameters provide AWR routing with intelligence to use wireless links based on conditions suitable for radio links. Using this feature allows AWR routing to balance the overall traffic in case of radio interference (Aruba Networks, 2011). For optimal video transmission, Aruba uses deep packet inspection, MAC layer frame prioritization, and buffering techniques to achieve optimal video transmission. Using the link layer metrics allows the network to use Air quality metrics which is specific to 49

50 wireless communications. Indicators for assessing current link quality include the link s data rate, received signal-strength indicator (RSSI) and external interference (Aruba, 2010, p. 5). The result is that AWR makes the routing decisions not just on network layer intelligence but adds layer-2 metrics for better routing decisions in wireless networks Aruba Networks Handoff process enhancement: Aruba Network uses MobileMatrix for mobile terminals moving from one subnet to another. The MobileMatrix takes approach similar to mobile IP, but uses the AWR fast routing convergence which helps alleviate the latency inherent in the mobile IP based handoff (Aruba, 2010, p. 10). When a handoff occurs the MobileMatrix roaming allows fast convergence for routing table updates. The location data is maintained in the routing table with each node. For mobile node position change, the routing tables are updated. The convergence happens faster than traditional layer-3 convergence, due to proprietary AWR routing Cisco s mesh network deployments: Cisco uses three components for deploying the wireless mesh network: 1. Cisco 1500 Series mesh AP: This is the access point for the wireless clients. 2. Cisco Wireless LAN Controller (WLC): WLC is the central point for controlling APs. 3. Cisco Wireless Control System (WCS) The Client mobile terminals connect to Cisco WMN with help of the Series 1500 mesh APs. The mesh APs are of two types, the Roof-top APs RAP) and mesh APs (MAP). Both the APs use dual radio channel for better bandwidth distribution between the client access channel (2.4 GHz) and backhaul channel (5.8 GHz). The WLC manage the mesh APs in that it dictates the security 50

51 policy, the QoS on the links and mobility functions.wcs in-turn centralizes the management of network management functions and dictates the RF prediction, network optimization functions (Cisco, 2009). MAPs use Adaptive Wireless Path Protocol (AWPP) to determine best path through other MAPs to WLC. There are two kinds of traffic on MAPs Bridge traffic: This is the traffic between the devices connected MAP Ethernet ports. Traffic between MAPs and WLC: This traffic carries WLAN traffic through the LWAPP tunnels Cisco s Light Weight Access Point Protocol (LWAPP): LWAPP is Cisco s underlying protocol used in communication between the MAPs and the Wireless LAN Controller (WLC) (Cisco, 2009, p.2-1). A MAP node establishes two types of tunnels between itself and the WLC. Both of these tunnels are LWAPP tunnels. The first tunnel is between MAP and WLC and carries the mobile client s layer-2 frame data, and the second tunnel carries the LWAPP control signaling traffic. LWAPP supports both layer-2 mode of operation where it transports layer-2 frame traffic between the MAP and WLC and layer-3 UDP traffic. The Layer-3 USP traffic mode of operation is more preferred as it allows network layer intelligence to be used in traffic distribution. 51

52 Figure 9: LWAPP tunnel setup Wireless LAN Controller (WLC) LWAPP Tunnel Traffic control messages LWAPP Tunnel Traffic W. Client Layer-2 traffic MAP/ RAP MAP/ RAP MAP/ RAP MAP/ RAP MT MT MT Mobility Management in Cisco wireless mesh networks Cisco defines a mobility group to group WLCs such that wireless clients can freely roam between WLCs in the same mobility group. If the client moves from attachment to a WLC A to another AP which maps to WLC B and both WLC A and B both belong to the same mobility group, then a layer-2 handoff is performed because the handoff involves same subnet. The movement of wireless clients across mobility group, which is across subnets, is defined as inter mobility group roaming. 52

53 Figure 10: Cisco mesh network mobility solution Mobility Group A Mobility Group B WLC WLC WLC WLC WLC WLC WLC WLC MAP Intra Mobility Group roaming (layer-2 handoff): To enable groups of WLCs to communicate with each other, Cisco defines a mobility group. A mobility group is a group of WLCs that together, act as a single virtual WLC by sharing essential information such as Radio resources, and VLAN parameters (Cisco, 2009, p.2-13). Any WLC in the mobility group can directly contact other WLAN in the same group using pre-authenticated tunnels. This enables the wireless mesh clients to freely roam between the 53

54 WLC which are part of the same mobility group without wireless clients having to reauthenticate. Cisco deployments use the mobility groups to facilitate seamless wireless mesh client roaming between APs that are joined to different WLCs which are part of same mobility group of WLCs. This way virtual WLAN domain is formed. The roaming within the mobility group is movement within a subnet. Thus this involves layer-2 roaming. WLCs which implement Cisco s LWAPP protocol use frame forwarding in layer-2 to exchange frames Inter Mobility Group roaming (layer-3 handoff): If the wireless client is moving from one subnet to another a layer-3 handoff is initiated. In this case an anchor/home WLC is defined and the anchor WLC becomes the gateway for the mobile client. The layer-3 handoff can be asymmetric or symmetric. In asymmetric handoff the path to wireless client is always via the anchor/home WLC and the path from the wireless client is directly to the foreign WLC. In symmetric layer-3 handoff both the path to and from the wireless client is via home/anchor WLC (Cisco, 2009, p.2-19). 54

55 Figure 11: Asymmetric layer-3 roaming Ethernet IP Tunnel Who is MT s WLC? MT s details, home IP.. Anchor WLC A-WLC Foreign WLC F-WLC Mobility Group S U B N E T A S U B N E T B MAP A MAP B MAP C MT MT 55