[Cu-wireless] 802.11 "mesh networking" standard

[David Young]

Interesting article here.  There are several useful 802.11 MAC
improvements this IEEE committee could make that are short of a
"full-blown mesh protocol" (which seems unlikely to me). Just for example,
a standardized way for measuring/reporting link quality would be useful
for "mesh" efforts.

Dave

<http://www.iapplianceweb.com/story/OEG20031208S0068>

Intel, Cisco push new wireless mesh network spec

By Rick Merritt 
iApplianceWeb
(12/08/03, 03:24:58 PM EDT)

SAN JOSE, Calif. - Engineers from Cisco Systems Inc. and Intel Corp. are
kicking off an effort to standardize mesh networks, one of the hottest
new segments in wireless. The move comes as a handful of startups hit
the market with proprietary mesh-networking technologies embedded in
homegrown ASICs and software.

A study group for an 802.11-specific mesh standard will meet for the
first time at the next IEEE 802 meeting in Vancouver, British Columbia,
Jan. 12-16, aiming to open the door to a broad class of interoperable
meshes. Such networks are leading candidates for a last-mile broadband
solution, said Peter Ecclesine, a technology analyst at Cisco who is
helping to organize the group.

Some analysts and industry observers applauded news of the move, but
several startups rolling out their own mesh approaches, including BelAir
Networks (Kanata, Ontario), Firetide (Los Gatos, Calif.) and Strix Systems
(Westlake Village, Calif.), expressed doubts that such a standard will
succeed.

Wireless meshes, typically self-configuring ad hoc networks of 802.11
access points, have become a hot topic with the rapid rise of wireless
local-area networks. Businesses and public hotspot operators consider
meshes to be quick and easy ways for both to stitch together many small
802.11 networks into much larger ones.

But the mesh-network products coming to market from more than half a dozen
companies are not interoperable. "My fear is that over time this will lead
to a fragmented market," said Steve Conner, a wireless R&D engineer in
Intel's corporate labs, who is helping to organize the standards effort.

The standard could also deliver a more efficient means of mesh networking
for 802.11 nets than currently exists, he added. "The 802.11 protocol was
not designed with mesh as a primary use, the MAC [media-access controller]
is inefficient across multiple hops and my fear is that most meshes
would be inefficient and waste the spectrum available," Conner said.

A standard could define a way to capture and control network statistics
to assist mesh links or even provide a full-blown mesh protocol. Although
the Internet Engineering Task Force is developing a number of mesh routing
protocols, none take into account the vagaries of wireless links. That
may not be suitable for the 802.11 standard, Conner said.

The January meeting will be the first attempt by a study group to
define what the standard should address as part of a subsequent IEEE
task group. The entire effort could take three years, said Conner,
who is developing a mesh protocol at Intel that could span 802.11 and
ultrawideband networks.

The study group is likely to consider a March proposal from the Navy
Research Laboratory that suggested an extension to 802.11 for mesh
networking. The Navy proposal defined an algorithm it prototyped that,
in less than half an Mbyte of open-source software, runs a mesh of 32
access points from multiple vendors.

The Navy based its work on adding some over-the-air messages to the 802.11
standard at what analyst Ecclesine described as Layer 2-1/2. The protocol
extensions would effectively extend the 802.11f work on communications
between two access points to include comms among multiple access points.

The Army is already using mesh technology, Ecclesine said. "Tanks in the
Iraq war had communications such that with no single point of failure
each tank could see the sensor data from any other tank," he said.

Work is already afoot in 802.11 to define standards around meshes that
could open the door to smart highways where cars communicate with one
another automatically. Ultimately, such meshes will be a key in wireless
broadband links that form the last mile to the home, Ecclesine said.

Many wireless standards have already defined some mesh capabilities. The
current work is specific to 802.11 but could spin off to cover other
wireless nets such as ultrawideband, Ecclesine added.

"It's not too early to be starting a standards effort in this
area," said Craig Mathias, a consultant for Farpoint Group (Ashland,
Mass.). "Meshes represent probably the next great architecture going
forward in wireless. They solve problems in availability, reliability,
load balancing, performance and throughput, though they do exact a toll
in power."

"A standard for meshes is a great idea," agreed Roger Durand, director
of RF systems architecture for Propagate Networks (Acton, Mass.), a
developer of software that enables mesh networking on Atheros 802.11 chip
sets. "Mesh networking is great for anything mobile on a large scale,
like emergency services or smart highways-things some people say are
still pie-in-the-sky but are really just over the next hill."

One of the first startups in the field, MeshNetworks Inc. (Maitland,
Fla.), is working with the standards effort in the hope that it could
supply some of its routing and automatic power control technology to
the initiative.

"We are transitioning from a product to a licensing business model," said
Rick Rotondo, vice president of technical marketing at MeshNetworks. "We
think we could contribute some significant IP to the standard and still
have plenty of value-added left over beyond that."

MeshNetworks already has one unannounced licensee and hopes to land a
second by February; both are large telecom companies. The startup, which
to date has sold systems supporting more than 30 networks worldwide,
recently closed a venture round of more than $10 million that included
an investment from Motorola Inc.

Some other newcomers in the field disagreed about the need for a mesh
standard. "It sounds good, but it's not so easy," said Bob Jordan,
vice president of marketing at Strix Systems.

Targeting business users, the startup in July started shipping small,
stackable 802.11 access points that form mesh networks. The company
said its systems use a discovery algorithm that constantly polls the
throughput, error rate, latency and other features of RF links to make
sure they are creating an optimal mesh while using only 1 percent of
the net's bandwidth for the polling.

"It's become a very interesting engineering problem to run a mesh, but if
you create a standard you risk coming up with a lowest-common-denominator
product, and performance will take a hit," Jordan said.

David Park, director of systems design at BelAir Networks, agreed. The
company has developed its own enhanced 802.11a chip sets and software to
create access points that use three radios to provide, on a point-to-point
backhaul link, 54 Mbits/second at ranges from one to 10 miles. The
systems-designed for large service providers, apartment buildings,
convention centers and airports-are now in trials with several North
American users.

Tropos Networks Inc. (San Mateo, Calif.) has a similar aim of creating
wide-area 802.11 networks based on proprietary mesh technology embedded
in its access points. Like some of his competitors, Bert Williams,
the startup's vice president of marketing, took a skeptical view of the
standards effort.

"If you standardize in the mesh itself, you'll stifle innovation,
dictating features by committee rather than letting vendors go off and
provide optimal performance," Williams said.

"The real use for a standard is when it comes to connecting different
clients to the mesh or linking the mesh to wireline routers," he added.

Ike Nassi, founder and chief technology officer at startup Firetide,
said trying to create a mesh-networking standard "is like trying to mix
and match ports from Cisco and Nortel switches: It's never been done."

On Jan. 5, 2004, Fireside will launch its HotPoint 1000S wireless mesh
router, which creates an 802.11b mesh linking multiple off-the-shelf
access points. The system uses proprietary software based on technology
developed by SRI that makes a mesh network appear like a multiport
Ethernet switch.

The Firetide software essentially puts a wrapper around incoming packets
to store information such as a client's MAC address. The code runs on
an embedded X86 single-board computer.

"It's true we don't interoperate with other mesh networks, but there
are a lot of things we do that that the other guys don't to make mesh
networking easy to use," Nassi said.



-- 
David Young             OJC Technologies
dyoung at ojctech.com      Urbana, IL * (217) 278-3933