Mesh nets offer challenges

Though a late entry, Cisco’s first wireless LAN mesh product will throw fuel on an already hot market.

Wireless mesh networks based on the IEEE 802.11 standards offer a range of benefits, such as fast deployment, minimal cabling, unlicensed spectrum and broadband throughput rates over wide areas. Mesh promises to make 802.11-based data and voice services pervasive and affordable, at least in campus and metropolitan-area deployments.

Those benefits are persuading a variety of customers to adopt mesh topologies: police and fire departments; campuses; construction companies; utilities; and, increasingly, local governments for business and residential Internet access.

Cisco is throwing the Aironet 1500 into a highly competitive and lucrative market. So far, a pack of upstarts, including BelAir, Firetide, Strix and Tropos Networks, and one established vendor, Nortel, have had the market to themselves, offering an array of architectures and technologies. Other companies such as Motorola, which acquired MeshNetworks, offer proprietary wireless mesh products. The 1500 makes Cisco a major player in this market.

“It appears to be a very competitive product,” says Craig Mathias, principal with Farpoint Group, a wireless consulting company. “And it does have ‘Cisco’ on the front label.”

A study from ABI Research estimates the global market for 802.11 mesh products to be $116 million and that it will soar to $1.3 billion five years from now. Much of that will be spent by municipalities or service providers to offer low-cost wireless Internet access, says Sam Lucero, ABI senior analyst.

“Mesh is attractive because of its [low] cost and ease of deployment,” Lucero says. Outdoor mesh access points cost $2,000 to $6,000, with an average price of $3,000; roughly 20 nodes are needed to cover 1 square mile, he says.

“But this is much more cost-effective than trying to trench new fiber or cable along the streets,” he says. “And in some cases, [mesh] is the only way to supply these services.” Laying fiber can cost up to $150,000 per linear mile, according to Lucero.

Mesh networks can start with just the barest number of nodes needed to blanket a given area, Mathias says. As users and traffic increase, more nodes can be installed. Costs are held down because you don’t have to provide a wired backhaul connection to each node: traffic from clients can hop through several wireless nodes (which sometimes can draw power from a battery or solar cell without needing an electrical connection) to reach a gateway that links with the wired network.

But there is very little that’s standard in mesh products.

“There are so many differences in products and approaches,” Mathias says. “The secret sauce is the mesh-routing protocol. And everyone implements them differently. Under what conditions do you hop [from one node to another], and to which nodes? How do you actually route packets through the mesh? All these are answered differently.”

Some products, notably from Tropos, use a single radio to handle client and backhaul connections. Cisco and others use separate radios for each connection. Some vendors will use two or more radios for both tasks. “Should you use multiple radios or not? And if yes, how many?” Mathias asks. “No one to my knowledge has done comprehensive tests of these products . . . to give us comprehensive benchmarks.”

The department of electrical engineering and computer sciences at the University of California at Berkeley is a beta site for the Aironet 1500. The department has 10 nodes deployed around the university’s campus, which has hundreds of buildings.

This deployment proved an unexpected challenge, says network manager Fred Archibald. Access points are designed for mounting on the horizontal arms of municipal light poles. But few of these were on campus. “We mounted them on roofs, about 70 to 90 feet above the ground,” Archibald says. “We had to use specialized directional antennas to get the signal back down to the ground.”

There was more interference than expected in the 802.11 5-GHz band Cisco uses to link the nodes, Archibald says. The interference was apparently caused by the large number of wireless research projects at the engineering school, he says.

The 1500 nodes are divided into groups, each of which is assigned its own radio channel, optimizing throughput for users and nodes in each group. But that means fewer routing options for the nodes in each group in case of outages or interference, Archibald says. “The mesh is not as flexible in reconfiguring itself,” he says. “Cisco has told me they will get better in this in the future.”

Troubleshooting is more complicated in a mesh. “The route [for a client’s packets] is more complex, and it’s constantly changing,” he says. One major upgrade during the beta test was the introduction of a GUI for the network-monitoring software. The display can now show a network administrator the path that a client packet has taken through a mesh.

Finally, the beta test revealed wide differences in how 802.11 client adapters cope with variable signal-to-noise ratios (SNR) that can occur outdoors. “Some don’t respond if the SNR is not as ‘friendly’ outdoors,” Archibald says. “Some are rock solid.”

Mathias doesn’t expect mesh networks to take over in carpeted offices. “But there are plenty of cases where you don’t have any wiring, such as loading docks or warehouses; concrete, not carpeted sites,” he says, “and campus nets where you have a hybrid of outdoor and indoor wireless.”