May 09, 2008, 2:15 PM — The proposed IEEE wireless
LAN mesh specification is already getting some traction, though still over
18 months from final ratification, thanks to early experimentation by the One
Laptop Per Child Foundation and a recently launched open source project.
This hands-on experience, fed back into the work of the IEEE 802.11s Task Group,
has already led to several changes in the draft standard, with other changes
being considered. When deployed, 802.11s will let different types and brands
of wireless devices find each other, interconnect securely, and forward traffic
on behalf of other mesh nodes, forging new paths automatically if nodes move
or wireless links fail. WLAN deployments will be cheaper and easier, say advocates.
PacketHop
earlier this month announced it will release in July the first commercial implementation
of the IEEE mesh draft .
The 802.11s standard promises to cut wireless networking loose from its Ethernet
cable moorings, creating what some call "opportunistic" networks.
"If you think about this underlying technology to form opportunistic connections,
it changes a lot of the assumptions of wireless LANs," says Robert Withrow,
adviser to the CTO Common Engineering Group, part of the Office of the Chief
Technology Office for Nortel Networks, which supports both the OLPC and the
open80211s
project, and was an early and active member of the 802.11s task group, which
launched in 2004.
"A piece of information might be coming from your [mesh] neighbor,"
Withrow says. "Well, who is that neighbor? Does he have a right to give
you information?" He also says that wireless networks are rapidly moving
from people-to-people connections, to people-to-machine and machine-to-machine
connections, which are vastly more numerous, dubbed by Nortel
as "hyperconnectivity." "It's difficult to do this in a hierarchical
way," he says.
The 802.11 standard does specify what's called an ad hoc mode, which lets,
for example, wireless laptops connect with each other. But this requires every
participating node to be connected directly to every other node. An 802.11s
mesh doesn't have this limitation. Proprietary mesh protocols abound, mostly
for outdoor wireless networks, though Ruckus Wireless and Aerohive recently
unveiled indoor mesh gear. But these protocols are for mesh infrastructures,
not clients, and because they're proprietary, they don't interoperate. (Compare
wireless mesh products with our online Buyer's Guide.)
OLPC embraces mesh
The OLPC began studying mesh benefits in 2006. Mesh
would let the group's inexpensive Linux-based
laptops create their own wireless network and communicate, without the need
for access points or intermediate servers, as well as share whatever wide-area
connection, such as a satellite link, that might be available for a village
school in Cambodia, Brazil, or Haiti.
Using the nascent 802.11s draft was a natural choice, says Michail Bletsas,
chief connectivity officer for OLPC. The 802.11s mesh works at Layer 2, so no
changes would have to be made to the TCP/IP network stack or other higher layer
applications.
One corollary benefit of that, he saw, was that the 802.11s code could be run
on the 802.11 network adapter module, with a system-on-a-chip having its own
memory and small CPU. That meant the OLPC laptop could suspend or shut down
its main CPU to save power, but trickle some juice to keep the radio alive as
a mesh node, forwarding traffic on behalf of other mesh participants.
"It only takes 800 milliwatts, compared to about 12 watts if it's implemented
on the laptop," Bletsas says. With some additional optimizations, OLPC
can cut the power needed to 400 milliwatts in some cases.
