WLAN Architectures and IEEE 802.11s Abstract WLAN mesh networks using 802.11 Access Points (APs) to relay traffic among each other to increase the RF coverage of the networks are becoming very popular for implementing wireless municipal networks in the past few years such as the infamous Philadelphia s city-wide WLAN mesh network. Depending on the wireless coverage and the expected traffic usages of the city/municipal, it is not uncommon to have a WLAN mesh network to employ a few hundreds or even ten of thousands of APs to cover areas in hundreds or even thousands km 2. Many network equipment vendors and start-ups offer 802.11 WLAN mesh products using proprietary technology and protocols. In order to increase the acceptance of the WLAN mesh technology, IEEE has formed the 802.11s workgroup to solve the mesh equipment interoperability issue by defining an architecture and protocols that support both broadcast/multicast and unicast delivery using radio-aware metrics over self-configuring multihop topologies (i.e., mesh network). With the agreement between the two competing mesh proposals from Wi- and SEE in the 802.11s workgroup in early 2006, a draft 802.11s proposal is expected to be available in 2007. This draft proposal is expected becoming an IEEE standard by 2008. This paper introduces the operations of contemporary WLAN mesh networks and highlights the areas that are being worked on or standardized by the IEEE 802.11s workgroup. Important WLAN mesh equipment selection criteria and network engineering parameters that are applicable for the current and the future 802.11s complied WLAN mesh systems and networks are discussed. Status: Draft Page 1 of 14
1 Introduction The Wireless Local Area Network (WLAN) mesh technology offers low-cost and high bandwidth wireless access and is being adopted by many municipalities/cities around the world for implementing wireless municipal networks. Traffic-light camera video streaming, electrical smart-meter telemetry, and public outdoor Internet access are some of the common network applications running over the WLAN mesh networks. The following diagram shows a typical WLAN mesh network setup. Contemporary WLAN Network Setup wireless link (802.11a or WiMAX) Backend servers such as DHCP, RADIUS, Payment, NMS, SIP & Captive portal servers AP1 Secured backhaul tunnel Portal L2 or L3 network wireless link (802.11a or WiMAX) Security and Roaming WLAN Controller WLAN device WLAN device Access wireless link 802.11b/g AP2 To other APs VLAN or IP subnet traffic Intranet or Internet Legend 802.11b/g (2.4GHz) access antenna antenna Status: Draft Page 2 of 14
Both the APs and the Portal provide wireless access and backhaul traffic relaying functions. However, the Portal has a wired connection (e.g., Ethernet) to the backhaul network and it is responsible for aggregating traffic for multiple APs (usually 3 to 10 APs) to form a mesh community. In the above diagram, the AP2 uses its wireless backhaul link to forward its access traffic from the WLAN devices such as wireless laptops or traffic-light video cameras to the Intranet or Internet via the AP1 and the Portal. The backhaul wireless links among the APs are usually WiMAX or 802.11a based. When 802.11a (5GHz) technology is used for implementing the wireless backhaul links among the APs, the APs will provide only 802.11b/g (2.4GHz) wireless access service for the WLAN devices so that sufficient frequency separation is available between the access and backhaul radios inside the AP to minimize self-interference. Many WLAN mesh products utilize 802.11a OFDM modulation for implementing the wireless backhaul links. It is found that OFDM performs reasonable well for non-line-of-sight (NLOS) communication among the APs to allow more flexible AP deployment. It is likely that mesh equipment vendors will take advantage of the newer modulation scheme Scalable Orthogonal Frequency Division Multiplexing Access (SOFMDA) with sub-channelization offered in the IEEE 802.16e standard to offer improved NLOS communication among the APs when the cost of the 802.16e radio is reduced. The typical separation distance between two APs is about a few hundred meters in metropolitan mesh network deployment. In a rural mesh network deployment where access traffic density is lower, the APs can space further apart so that less number of the APs are required to cover a given rural area. Most APs support the use of different external backhaul antennas for different deployment scenarios. For example, with a typical AP s backhaul radio outputting at 26dbm (400mW), a higher gain (e.g., 8dbi) backhaul antenna can easily allow two APs to space over 1km in rural mesh network deployment while meeting the regulator s requirements (e.g., EIRP of less than 36dbm or 4W in higher 5GHz UNI band in Canada). Most of the current WLAN mesh solutions employ centralized WLAN Controller(s) for securely terminating mesh network s traffic to and from the APs. The WLAN Controllers also offer subscriber mobility by allowing subscribers to roam among the APs without session drop or user re-authentication. Depending on the design of the WLAN Controller, one WLAN Controller can usually support a few hundreds APs. Usually, the WLAN Controllers offer option for cascading them together to support a larger WLAN mesh network that requires thousands of APs. Since the high-level design and architecture of WLAN mesh networks are very similar among different mesh equipment vendors, the vendors attempt to differentiate their products in the following areas: mesh routing protocols traffic forwarding optimization and resilience antenna design (MIMO, beamforming) wireless distance, and coverage backhaul radio design - multimedia traffic handling capability Status: Draft Page 3 of 14
Besides the APs, Portals, and the WLAN controllers, a WLAN mesh network needs the following backend servers for proper operations: DHCP server - subscribers and network devices IP address assignment Captive Portal server - WEB user authentication for public Internet access RADIUS server - user and network device authentication Payment server network billing and payment such as on-line credit card authentication SIP server - voice over /IP operations NMS WLAN mesh network management Firewall real-time virus scanning and network attack The above lists the commonly used backend servers for a WLAN mesh network. Other servers such as content cache servers may require depending on the requirements of the WLAN mesh network. Almost all WLAN mesh equipment vendors claim that their products can be used to build a scalable, secured, cost-effective, and resilient wireless network that requires minimum configuration effort such as: Topology Discovery and Channel Allocation to minimize configuration effort Path Selection and Forwarding for optimal and alternative traffic paths under network failure Access Traffic Security and Subscriber Authentication for secured access traffic protection Traffic Security and Device Authentication for secured backhaul traffic protection Traffic management for multimedia wireless access traffic Network management for ease of WLAN mesh network operations and maintenance Coincidentally, these are also the areas and functions that the 802.11s workgroup will address and standardize for mesh device interoperability. The following sections describe the operations of each of the above functions in a contemporary WLAN mesh network. The information is useful for understanding the upcoming 802.11s draft and standard when they are available in 2007/08. 1.1 Topology Discovery and Channel Allocation When a AP is powered up, it will scan its backhaul frequency channels to try to form backhaul wireless links with its adjacent APs. The ultimate goal is to locate its corresponding Port for access traffic forwarding to the Intranet or Internet. Many WLAN mesh products utilize the unlicensed 5GHz band for implementing the wireless backhaul links. Since the unlicensed 5GHz band is not harmonized in the world, different 5GHz UNI bands are being used in different countries. For example, in Canada, the unlicensed upper UNI band from 5.725GHz to 5.825GHz is usually used by WLAN mesh vendors for Status: Draft Page 4 of 14
implementing the wireless backhaul links. There are altogether six non-overlapping channels within this spectrum under the 802.11a modulation. In general, a AP will only form two to three wireless backhaul links with its adjacent APs for performance reason. The discovery protocols exchanged by the APs use many different parameters for determining which backhaul wireless links should be formed among the APs such as: Relative Signal Strength Index (RSSI) Loading on the adjacent APs Signal to noise ratio Interference on the backhaul channel frequency Network redundancy and topology optimization Packet errors Depending on the number of backhaul radios (physical or virtual) and the type of backhaul antennas (omni-directional or directional) are used at the APs, the mesh network topology discovery and channel allocation protocols and processes can be fairly complex. Most mesh equipment vendors use proprietary network discovery protocols to achieve minimum or no-effort WLAN mesh network configuration and expansion. 802.11b/g communication modes are normally supported by a AP for wireless access. There are eleven channels defined in the 802.11b/g 2.4GHz spectrum but only three of them (channels 1, 6, and 11) are non-overlapping in term of frequency bandwidth. Some mesh equipment vendors offer automatic access channel allocation with dynamic transmission power level adjustment to ensure that the APs provide adequate access link RF coverage for a given geographic area with minimum inter- AP interference. The continuous RF coverage ensures subscribers can roam among the APs without session drop or re-authentication. Pedestrian walking to slowly moving vehicle speed of 60km/h is normally supported by most mesh vendors as far as device roaming is concerned. Some mesh equipment vendors claim to support fast device handoff to support roaming speed of over 100km/h. The goals for access link channel allocation and dynamic power level adjustment are to: Maximum access channel reuse Minimum inter- AP interference Continuous access channel coverage for device roaming The following diagram illustrates an example for access link RF coverage to achieve the above requirements: Status: Draft Page 5 of 14
WLAN Access Link Channel Assignment and Dynamic Power Adjustment Ch 1 Ch 11 Ch 1 Ch 6 Ch 6 Ch 1 Ch 11 Ch 11 Legend Ch 1 Ch 6 Ch 1 Ch x Access Link RF coverage of a AP Control protocols such as CAPWAP are usually used to achieve automatic access link channel allocation and dynamic power level adjustment. 1.2 Path Selection and Forwarding A AP usually forms more than one wireless backhaul links with its adjacent APs for access traffic forwarding for network redundancy and load sharing purposes. Most of the contemporary WLAN mesh products use layer 3 IP routing for access traffic forwarding over multiple wireless backhaul links. This is also being adopted by the 802.11s workgroup instead of the layer 2 bridging mechanism. Similar to IP routers where they exchange IGP routing protocols such as OSPF or ISIS among each other to learn the optimal network topology under normal and failure conditions, APs also exchange routing protocols among each others over the control channel of the wireless backhaul links. Through the routing protocol exchange among the APs, the APs form the optimal network topology for relaying access traffic via the wireless backhaul links. When a network outage occurs due to equipment failure or RF interference, the mesh network routing protocol will update the forwarding table of the APs to forward access traffic via alternate paths over the mesh network. WLAN mesh network s routing protocols differ from the standard IP s IGP protocols due to the different characteristics of a WLAN mesh network such as link latency and reliability, and RF interference etc. There are more than 70 competing mesh routing schemes available such as the: Status: Draft Page 6 of 14
Ad-hoc On Demand Distance Vector Dynamic Source Routing Optimized Link State Routing protocol Temporally-Ordered Routing Algorithm Etc.. A few WLAN mesh equipment vendors champion their products around their proprietary mesh network routing protocols and claim that they can always find the optimal path for traffic forwarding. When access traffic arrives at the WLAN controller, the WLAN controller either tags the access packets with the designated VLAN ID or route the packets according to the IP subnet information. Usually, different user groups are being assigned to different SSIDs and their traffic are then segregated and forwarded based on either VLAN IDs or IP subnets at the egress ports of the WLAN Controller. There is not much difference between the VLAN and IP subnet forwarding designs in term of broadcast and multicast traffic containment. One of the key deliverables from the 802.11s workgroup is to standardize the AP routing protocol over a WLAN mesh network for device interoperability. 1.3 Access Traffic Security and Subscriber Authentication The existing WLAN access security and authentication mechanisms offered by the IEEE and the Wi-Fi Alliance are normally supported by a WLAN mesh network so that a user can use the same WLAN hardware and software for secured WLAN mesh access. WLAN security is a big alphabetical soup with many terminologies such as TKIP, WPA2, EAP- TTLS, EAP-MDS, and AES etc. For more information about WLAN security, please refer to the article WLAN Security and Wi-Fi Protected Access on IEC Annual Communication Review Volume 57. In general, the access traffic security of a WLAN mesh network is as good as any modern WLANs and wired enterprise LANs as long as the WLAN mesh network and devices are properly setup. 1.4 Traffic Security and Device Authentication As mentioned is the mesh network topology discovery section, when a AP is powered up, it will attempt to form wireless backhaul links with its adjacent APs. In order to ensure that it will not form adjacent links with rogue mesh APs to allow hackers from gaining illegal access to the WLAN mesh network, standard X.509 digital certificate with user-defined password are usually employed at the APs and the WLAN Controller for proper device authentication. Status: Draft Page 7 of 14
Once a AP forms adjacent wireless backhaul links with its neighbouring APs, it normally establishes a secured traffic tunnel from the AP to the WLAN Controller for securely transporting access traffic over the WLAN mesh network. Through digital key exchange, only the AP and the WLAN controller have the right pair of keys for encrypting and decrypting the traffic between the AP and the WLAN Controller. This secured traffic tunnel design has been adopted by many WLAN vendors. When evaluating existing WLAN mesh product, one should ask the equipment vendor how APs traffic is transported over the WLAN mesh network and is there any hardware accelerator inside the WLAN Controller to expedite the encryption and decryption processes, which are very CPU extensive. 1.5 Traffic Management Best-effort data over a large-scale WLAN mesh network involving thousands of APs has been implemented and proven in many large WLAN mesh network deployments. The next challenge for the mesh equipment vendors is to support high volume of multimedia traffic such as voice and video over a WLAN mesh network. This is becoming important as many handset vendors are releasing their 2 nd generation dual-radio handsets that support both Cellular (CDMA or GSM) and WLAN access. This 2 nd generation dual-radio WLAN handsets have the same appealing form factor as the latest 2G/3G handsets. The ability for a subscriber to roam between a 2G/3G cellular network (e.g., cover a larger geographical area at a higher connection cost) and a WLAN mesh network (e.g., smaller coverage area than cellular but offers higher data rates at lower connection costs) is very appealing to users. It is very likely that the 802.11s workgroup will adopt the IEEE 802.11e (WMM) standard for traffic management for WLAN mesh networks. This enables the APs to perform some sorts of traffic prioritization to meet the latency requirements of multimedia traffic. Note that WMM is simply a Class-of-Service (CoS) instead of a resource reserved Quality-of-Service (QoS) mechanism due to the limitation of the original 802.11 protocol. However, in light of the widely success 802.11 protocol and changing a new WLAN MAC protocol to offer better QoS is not possible in the near-term; WMM is a reasonable alternative for supporting multimedia traffic over a WLAN mesh network. 1.6 Network Management A WLAN network can comprise hundreds or thousands of APs each with multiple wireless backhaul links. Managing a WLAN mesh network of this size and complexity demands a powerful and intelligent network management system. One way to promote mesh device interoperability is to offer standardized mesh device and network SNMP MIBs so that operators can import various standard network MIBs into their network management system. Status: Draft Page 8 of 14
While the 802.11s workgroups will address many mesh device interoperability issues, the workgroup also deliberately leave out some WLAN mesh architecture designs for equipment vendors to differentiate their products. Two noticeable issues that are left out are the design of the backhaul radio (e.g., single vs. multiple radios) and interference mitigation in the unlicensed bands. Status: Draft Page 9 of 14
2 WLAN Network Design and Engineering Design and engineering a WLAN mesh network is very similar to any other wireless networks that involve site survey, equipment selection and evaluation, RF propagation simulation, and user traffic requirement gathering etc However, a WLAN mesh network has some additional attributes such as: Unlicensed spectrum and interference Many 3G carriers spend billions of dollars to secure the licensed 3G frequency spectrum in their countries for 3G cellular services. The fact that a WLAN mesh network uses the unlicensed 2.4GHz and 5GHz frequency spectrums is a double-edged sword. It makes deploying WLAN mesh networks fast and at a much lower capital costs. However, it also makes the network susceptible to innocent interference in these unlicensed frequency spectrums from devices such as microwave ovens and wireless home phone etc 802.11 MAC layer does not support QoS even with WMM Network engineering, planning, and management are more important in WLAN mesh networks to ensure that no mesh devices are overloaded during normal and peak operations to ensure adequate multimedia traffic support. The untrue expectation of a WLAN mesh network that it is self-configured and easy to scale with minimum engineering and management efforts The following summarizes some of the network engineering issues that are specific to WLAN mesh networks. They are applicable for the current as well as the future 802.11s complied WLAN mesh networks: Understand the current and/or projected traffic pattern, mobility requirements, traffic types (e.g., best effort data, voice over mesh etc ), wired and wireless (e.g., 3G cellular) interworking requirements, resilience strategy for discussion with the equipment vendors or network integrator Use of directional instead of omni-directional backhaul antenna to minimize interference in the backhaul frequency spectrum whenever possible Select WLAN mesh system that uses different backhaul channel for each wireless backhaul link for better interference mitigation (see the Appendix) Fully utilize the number of non-overlapping backhaul frequency channels among the APs when designing the WLAN mesh network Status: Draft Page 10 of 14
Perform a proper site survey to identify the locations for AP deployment and adjust the transmission power of the APs to minimize interference among the APs if the WLAN mesh system does not offer automatic AP power adjustment If the WLAN mesh system and the local regulatory allow a choice of operating frequency, select the spectrum that has less interference. For example, the 4.9GHz frequency spectrum used for US homeland security or the 3.5GHz for 802.16-2004 fixed point-to-point wireless link has less interference than the 5.8GHz unlicensed frequency spectrum Minimize or eliminate access traffic support at the Portal. This is because the Portal is responsible for aggregating traffic from multiple APs within a mesh community and it is usually the performance bottleneck of the system. When a single CPU is used inside a AP (see the Appendix AP Design), the less work for the Portal to process access traffic, the more CPU cycles it can spare for forwarding APs traffic If the WLAN mesh network plans to support a high volume of multimedia traffic such as voice and video, select APs with multiple physical backhaul radios for better throughput and latency performance thought it is generally more expensive Select a WLAN mesh system that has built-in access and backhaul WLAN intrusion and interference detection and containment functions Limit the mesh hop count to ensure that latency along the APs and the Portal is fell within the requirements of the multimedia traffic. For example, the one-way latency for voice over a WLAN mesh network should be less than 100ms A AP with multiple physical backhaul radios can support multimedia traffic better than a AP with only a single and/or multiple virtual backhaul radios. The industrial trend is to have multiple physical backhaul radios implemented inside a AP If the WLAN mesh system utilizes centralized WLAN Controller(s) for securely terminating mesh traffic, check with the vendors the scalability of the WLAN Controller(s) to make sure that multiple WLAN Controllers can be cascaded together to support a large WLAN mesh network. Also, ensure that the WLAN Controller has dedicated hardware for secured traffic encryption and decryption. Designing, installing, and managing a WLAN mesh network for a municipal with hundreds or even thousands of APs is not a small task. The promise by the mesh equipment vendors that WLAN mesh networks are self-configured without much manual intervention is far from the truth. Incumbent carriers with extensive experience and infrastructure in wired and wireless networks are in a better position to perform WLAN mesh network design, engineering, installation, and management. Status: Draft Page 11 of 14
3 Conclusion With the upcoming standardization of the WLAN mesh technology via the IEEE 802.11s workgroup, one can expect WLAN mesh networks being more acceptable by both carriers and users. Many municipalities around the world plan to operate their own WLAN mesh networks with the promise of ease of network expansion, lower cost, and minimum management efforts by the WLAN mesh equipment vendors. However, it has been found that design, install, and manage a WLAN mesh network is very similar to any existing carrier wireless networks and it is not as simple. Incumbent carriers with significant wired and wireless network experience and infrastructure are in a better position to design, engineering, install, and manage the WLAN mesh networks. This is especially true when the WLAN mesh network needs to interwork with the incumbent carrier s 2/3 G cellular network for device mobility. 4 Biography is a member of IET. He can be contacted at dcheung@theiet.org. Status: Draft Page 12 of 14
APPENDIX Since WLAN mesh network is normally operating in the unlicensed frequency spectrum, interference mitigation is an important design aspect of a WLAN mesh system. For the contemporary WLAN mesh products, there are two different backhaul frequency reuse designs that can affect how well the products mitigate interference: 1. One common backhaul channel for a mesh community Recall that a Portal is responsible for aggregating access traffic from multiple APs to form a mesh community. In this design, all wireless backhaul links within a mesh community comprising the Portal and the APs use the same backhaul frequency channel. Normally, omni-directional backhaul antenna is used in this design. One Channel for a Community Frequency A AP1 Portal L2 or L3 network Frequency A Frequency A AP2 AP3 Frequency A AP4 The advantage of this design is that the there is no switching latency between the wireless backhaul links when the AP has only one physical backhaul radio. This is good for implementing fast handoff and supporting multimedia traffic. The downside is that when there is RF interference on the backhaul frequency, all the APs and the Portal will have to change to a new backhaul channel and is likely to cause network downtime to the mobile subscribers. Status: Draft Page 13 of 14
2. Different backhaul frequency for each wireless backhaul link In this design, a AP selects different backhaul channels for each of its backhaul links with its adjacent APs. The APs intelligently utilize all the available backhaul channels for implementing the wireless backhaul links among them. This design is very good for interference migration because only one AP link need to change its operating frequency under RF interference. However, if a single physical backhaul radio is being used in the AP, the design will incur additional switching latency among the wireless backhaul links due to frequency changes of the backhaul radio. This may affect its support of multimedia traffic. Different Channel for Each Link Frequency A AP1 Portal L2 or L3 network Frequency C Frequency B AP2 AP3 Frequency A AP4 Status: Draft Page 14 of 14