A costly and complex aspect of today's wireless networks can sometimes be the very component they're supposed to eliminate: cabling. Emerging 802.11-based mesh networks attempt to resolve this irony by using more radio spectrum and less wire in the form of Ethernet cabling than traditional wireless LANs. These are early days for WLAN meshes, but proprietary infrastructure products are commercially available. Organizations with difficult-to-cable environments and those that frequently move their WLAN nodes are among mesh's early adopters.
A wireless mesh infrastructure is, in effect, a router network minus the cabling between nodes -- with the inherent rerouting for fault tolerance that such networks deliver. It's built of peer radio devices that don't each have to be cabled to a wired port like traditional WLAN access points (AP) do. Rather, each simply plugs into an AC power supply. It automatically self-configures and communicates with other nodes over the air to determine the most efficient multihop transmission path.
Today, the way these functions work is unique to each vendor. So enterprises that build mesh networks will likely use one vendor for a few years until standards are in place.
"Mesh is a reasonably important enterprise architecture going forward, because it dramatically simplifies installation," says Craig Mathias, a principal at Farpoint Group, a consulting firm in Ashland, Mass. "You take a node out of the box, plug it into the wall -- end of discussion."
Supplying power to a mesh node can still be problematic. However, electrical outlets are usually far more abundant in buildings than Ethernet ports are, Mathias notes.
Only devices at the very edge of the wireless mesh hit wire -- either to connect to a network switch or to stand-alone cabled devices such as printers and video cameras.
A design goal is to minimize the number of those wired devices and allow network managers to easily move wireless nodes as needed for capacity and coverage.
In a wireless mesh network, as devices are added and moved, the network automatically discovers topology changes and adjusts traffic-forwarding paths to optimize throughput.
Urology Clinics of North Texas replaced a traditional WLAN with a meshed Access/One system from Strix Systems Inc. in Calabasas, Calif., for just this reason. "We had intermittent problems with interference and shifting coverage holes," explains Kyle Nash, IT manager at the Dallas-based facility. This required him to frequently move APs to tune the network, which was laborious and time-consuming because cabling ran from each AP to an Ethernet switch.
"Now I just move APs on the fly. This means the network is up longer. It will also make things much easier as our network continues to expand," says Nash, whose goal is for the five-office, 200-plus user facility to eventually be about 90 percent wireless.
Early Players and Users
The flexibility provided by mesh networks is particularly helpful over large geographies and in hard-to-wire buildings. Cisco Systems Inc., for one, says it helped kick off the effort to develop the IEEE 802.11s mesh networking standard when it discovered that some of its customers were running Cisco Aironet APs in "repeater mode," whereby one AP backhauls packets to another.
"This was happening in large warehouses where customers either couldn't get to a location or were running into Ethernet's 100-meter cabling limitation," says Jon Leary, product line manager in Cisco's wireless networking business unit.
Similarly, consider hospitals using the services of Shared P.E.T. Imaging LLC in Canton, Ohio. The company offers mobile positron emission tomography (PET) diagnostic medical imaging services to facilities that can't support the service full time in-house.
Mobile scanning labs are equipped with an US$800 Firetide Inc. 802.11b HotPoint mesh router attached to a PET scanner. The router in the mobile coach communicates images to another router inside the hospital, where they are relayed to a reader, says Marc Simms, director of IT at Shared P.E.T. Previously, the company dragged Category 5 Ethernet cabling outdoors after drilling a hole in the building.
Simms describes the cabling as "flaky and susceptible to weather." In one instance, cabling beyond the 100-meter Ethernet limit required the installation of more-costly fiber optics.
Simms says that before the company took the wireless mesh approach, an installation with a new customer took four to eight weeks and cost $2,000 to $4,000 -- or $10,000, if fiber was involved. "Now, setup time is about an hour," he says.
Strix and Los Gatos, Calif.-based Firetide are the two mesh vendors that have made the greatest enterprise inroads. Firetide focuses strictly on wireless backbone applications -- the company added 54Mbit/sec. 802.11a nodes to its portfolio this month -- while Strix builds nodes that perform double duty as wireless backbone routers and traditional WLAN APs.
Strix supports 802.11a/b/g in a modular, stackable mode that costs $800 to $900. It also uses the faster, shared 54Mbit/sec. 802.11a or g for backhaul and 802.11b for user access. In addition, Strix says it can use the proprietary 802.11g channel-bonding mode supported in some WLAN chips to achieve 108Mbit/sec. optimum-shared bandwidth.
Like Strix and Firetide, Tropos Networks Inc. in San Mateo, Calif., makes both indoor and outdoor Wi-Fi mesh products that could be used by enterprises or public network operators. To date, though, Tropos products have been installed in metropolitan applications, such as citywide 802.11 hot-spot networks.
Similarly, Nortel Networks Corp., which began shipping mesh products last month along with BelAir Networks Inc. and RoamAD, focuses on outdoor applications such as municipal and campus backbones. Like Strix, Nortel mesh nodes also support traditional AP access, and Kanata, Ontario-based BelAir's products offer indoor coverage from the same, outdoor-mounted node. Nortel provides indoor WLANs via a product line from Airespace Inc., which it resells.
The enterprise applications for mesh are fairly targeted, given the relatively low speeds of Wi-Fi networks compared with gigabit-speed cabled Ethernet backbones. The actual throughput speeds of Wi-Fi are about one-half to two-thirds of their stated optimum bandwidth because of wireless overhead and interference.
Generally, adding more mesh nodes increases capacity. However, Wi-Fi bandwidth is shared, and while both Strix and Firetide cite per-hop latency of less than 5 msec, this could add up as meshes scale, particularly as voice applications emerge and more hops eat into the total voice-latency budget.
"The concept of sustained (wireless mesh) backbone bandwidth is not applicable," says Sunil Dhar, director of product management at Firetide. "End-to-end throughput is determined by the number of wireless hops required to traverse the mesh, the density of the mesh deployment and the amount of interference."
"For backhaul, you're going to pay, performance-wise," says Yuval Goren, a wireless consultant at Goren International in Saratoga, Calif. "Mesh is for applications where you can't get access wherever you want, such as temporary applications where it makes no sense to pay thousands of dollars to run cables."
The Computer History Museum in Mountain View, Calif., has such an application. When the previous occupant vacated the building a few years ago, it cut much of the Ethernet cabling out of the 120,000-square-foot building.
So the museum is using Firetide nodes for temporary exhibits and outdoor events, where generator power is used, and in small, inaccessible areas indoors, according to Mike Walton, the museum's director of IT.
The Computer History Museum runs a Hewlett-Packard Co. AP infrastructure hanging off the Firetide 11Mbit/sec. backbone, which serves places such as the lobby for event registration.
"The lobby had no options for an Ethernet drop. Now, if I want registration computers, all I have to do is plug a (Firetide) node into an electrical outlet," Walton says.
Similarly, The Science Place, an interactive museum in Dallas, has exhibits requiring both LAN and Internet access that are frequently moved. The 38,000-square-foot, two-building facility has 20-foot ceilings, 6-foot limestone walls and strict laws about what can be done to the building, says Michael Wright, director of IT.
The Strix network runs through both buildings "without having to string wires across the floor," Wright says. However, the organization retains a wired Fast Ethernet backbone with 10Mbit/sec. throughput to the desktop for permanent administrative uses.
Jose Villarreal, director of technical marketing at Strix, counters that it won't be long before wireless is on par with wired speeds. "Some seriously smart engineers used to say we would never be able to deliver broadband over twisted pair. How many DSL deployments are being done each day now?"
Wexler is a freelance writer based in Campbell, Calif. You can reach her at email@example.com.
Standards on the Way . . . In 2007
The IEEE 802.11 Task Group S met in September to begin developing a standard for interoperable wireless LAN mesh infrastructures. There are at least two items that the standard will likely define: 1 How a packet selects its multihop path across the wireless mesh. Today, this is the secret sauce giving early vendors their value-add. Best-path selection is analogous to traditional Layer 3 wired routing protocols, such as Open Shortest Path First, or OSPF. In wireless configurations, though, the algorithm must be tightly coupled with Layer 1 radio metrics, accounting for physical-layer issues such as signal strength and interference when selecting a path. 2 Authentication and security extensions for mesh. Part of the standard's scope is still to be decided, says Steven Conner, wireless network architect at Intel Corp.'s Communications Technology Lab and technical editor for the IEEE 802.11 Task Group S. For example, he speculates that the standard could include the way a node selects a channel and what transmission power to use. Conner predicts that proposals will be received during the first half of 2005 and that the standard will be complete sometime in 2007.
The term mesh network refers to a network topology in which devices are connected with many redundant interconnections between network nodes. In a true mesh network -- wired or wireless -- every node has a connection to every other node in the network, whether directly or via hopping though intermediate nodes.
Looking specifically at wireless mesh networks, they are actually more truly switched wireless networks than those that use so-called wireless LAN switches.
In a mesh network, packets are forwarded wirelessly from node to node. Products called WLAN switches, on the other hand, don't actually support any wireless interfaces.
Note, though, that mesh architectures don't preclude the use of a WLAN switch or other centralized management architecture on the wired side of the network. -- Computerworld (US)
Join the CIO New Zealand group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.