Funny thing about wireless networks, they demand a heck of a lot of wires. All those access points that extend your network over the air are connected by wires to the Ethernet LAN.

In a building with Category 5 already running everywhere or in a confined location like a coffee-shop hot spot, the wiring issue might not be such a big deal, but what happens when you're faced with a setting where stringing wires is impractical? Or what if you need to reconfigure an existing wireless network on the fly, or if you need to set up a new network in response to a time-specific event? Or what if you want to provide Wi-Fi to a college or corporate campus, or to a downtown area?

In these cases, wireless mesh technology might be the answer. The IEEE has acknowledge this, setting up 802.11s to look at the issue.

The Defense Advanced Research Projects Agency (DARPA) began studying mesh technology in the mid-1990s for use in battlefield communications. Soon after, research-and-development companies such as SRI International began looking at wireless meshes, and now a number of start-ups say this technology will help them transform the wireless landscape.

Simply put, wireless meshes use a topology of redundant connections between nodes to create a self-configuring, self-healing network. Because very few nodes need to be wired into the network, mesh networks limit the need for expensive backhaul while providing additional benefits such as dynamic optimized routing and automatic load balancing.

Client mesh vs. infrastructure mesh?
Not all mesh topologies are created equal - some include client devices in the mesh, but most don't.

In a pure client mesh, every device in the network, including laptops, PDAs and smart phones, can pass along traffic for other devices, which lets users "hop" through neighboring devices or wireless routers to communicate with each other and to reach the wired network. This "multi-hopping" capability creates a meshed network that automatically routes around congestion and line-of-sight obstacles, while improving throughput as more clients enter the network.

In an infrastructure mesh, access points and wireless routers carry the traffic back to the wired network. In the infrastructure camp, you'll find Tropos Networks, Firetide and BelAir Networks.

"In our system, the clients don't participate in the mesh at all," says Bert Williams, vice president of marketing for Tropos Networks. "They attach to the mesh, but the meshing itself is done by infrastructure devices."

MeshNetworks and PacketHop let clients actively participate in the mesh, but it's not accurate to call either of them pure client-mesh companies. Instead, both rely on infrastructure nodes for the core of the mesh network. Then, once a mesh infrastructure of access points or routers is in place, clients can be used to extend the coverage area, fill coverage gaps and communicate directly with nearby devices.

... or fixed mesh versus mobile?
"I would actually split the mesh world up differently," says Michael Howse, CEO of PacketHop. "Rather than client vs. infrastructure meshes, what I see are fixed vs. mobile meshes. The fixed-mesh companies string together access points. There is one wired node that provides backhaul, with other wireless nodes on the same subnet." There are benefits to this, Howse says. The wireless-mesh nodes provide better coverage, while also distributing intelligence. "The problem with this model is that you will compete with Cisco, Proxim, Symbol and other access-point providers."

Fixed-mesh is a cable-replacement strategy. For example, Strix Systems says its main competition comes from cable installers. "Our mesh technology allows for deployments of (wireless) LANs in locations where Ethernet is difficult, expensive or even impossible to run," says Bob Jordan, vice president of marketing for Strix.

Jordan says Strix recently unwired the Central Texas Parole Violator Facility in San Antonio. The prison was concerned about the difficulty of installing Ethernet throughout the facility. With the Strix solution, a few backhaul nodes support a large network of wireless, access-point-like nodes that can be placed anywhere there is a power outlet.

Likewise, Bert Williams, vice president of marketing for Tropos, says the cable-replacement strategy offers the most near-term opportunity. "When you look at a traditional WLAN deployment," he says, "you'll find that 90 percent of the installation cost is directly related to wiring."

PacketHop's Howse argues that fixed-mesh companies lack mobility. In a fixed mesh, clients lose sessions and must re-associate as they move from one access point to another.

Peter Stanforth, CTO of MeshNetworks, says much of the problem with fixed meshes is that they rely on outdated ideas from the world of wireline meshes.

The problem is twofold: First, the wireless medium is unreliable compared with a wired network. When you look at Internet routing protocols, for example, they make the assumption that the underlying physical layer is very stable, which is not the case with wireless. Second, mobile devices are constrained.

With a power-limited device, he says, you are always trying to solve problems relating to power, range and data rates. But you can solve only two of the three problems. "For a given power, I can either get high range or high throughput," Stanforth says. It's a classic trade-off, one that is being played out with 802.11 standards (802.11g has better range, while 802.11a has better throughput).

Mesh networks mitigate this problem, and the mobile mesh companies argue that end devices are the key to balancing the range vs. coverage debate.

Vendors also are taking different approaches in terms of whether their products are designed for indoor or outdoor use. For example, BelAir and Tropos are pitching "cellular Wi-Fi" for outdoor applications, while Strix is focusing on indoor deployments.

Mesh vs. WLANs
It would seem that mesh-network companies are on a collision course with WLAN vendors.

For example, the city of Medford, Oregon, recently selected a wireless-mesh network over a WLAN to give public safety and public works agencies mobile access to voice, video, data and position-location services. "Most 802.11 wireless networks require the client to disassociate and re-associate with access points as you travel through a coverage area," says Doug Townsend, Medford's technology services director. "With a [wireless] mesh, there is no drop or delay typical in systems that require re-associating. Mesh [networks] also avoid the latency effect and decrease in bandwidth usually seen with most 802.11 wireless systems as you travel away from access points."

"If you look at the WLAN switch space, you have two approaches," Jordan says. "Companies like Trapeze and Airespace are pursuing thin access points and centralised intelligence, while Cisco believes in intelligent, fat access points. Mesh companies offer a third way, one that can coexist with whichever architecture you choose."

He says mesh networks provide a single point for managing the network, while, at the same time, pushing intelligence out into the network to enable optimal routing and automatic load balancing. Thus, if you have third-party access points and little intelligence in a network, meshing adds intelligence to wireless nodes. If you have fat access points, meshing offers a centralised point of control. The real difference, then, between the mesh and the various WLAN companies is the wires. "In many sectors, removing the wires is key. Consider emergency responders," Jordan says. "If you have a need for temporary workgroups and you set them up and tear them down all the time, is it really feasible to run wires?"

Adds Ike Nassi, CTO of Firetide Networks: "Our mesh solution simplifies installation. You plug it in to AC power, and you're done. Our mesh software is self-configuring and self-tuning. Besides limiting costly wires, you eliminate expensive site surveys as well."