[Previous entry: "Do Mesh Networks Scale? Another (biased) View"] [Next entry: "Ukraine's Snickers Bar"]

07/07/2004: "Even Microsoft does Mesh"

Microsoft Research appears to have gotten on the Community Wireless train.


Researchers in Microsoft Research Redmond, Cambridge, and Silicon Valley are working to create wireless technologies that allow neighbors to connect their home networks together. There are many advantages to enabling such connectivity and forming a community mesh network. For example, when enough neighbors cooperate and forward each others packets, they do not need to individually install an Internet "tap" (gateway) but instead can share faster, cost-effective Internet access via gateways that are distributed in their neighborhood. Packets dynamically find a route, hopping from one neighbor's node to another to reach the Internet through one of these gateways. Another advantage is that neighbors can cooperatively deploy backup technology and never have to worry about losing information due to a catastrophic disk failure. A third advantage is that this technology allows bits created locally to be used locally without having to go through a service provider and the Internet. Neighborhood community networks allow faster and easier dissemination of cached information that is relevant to the local community.


Source and binaries to their implementation are also available.

Microsoft Research has implemented ad-hoc routing and link quality measurement in a module that they call the Mesh Connectivity Layer (MCL). Architecturally, MCL is a loadable Microsoft Windows driver. It implements a virtual network adapter, so that to the rest of the system the ad-hoc network appears as an additional (virtual) network link. MCL routes using a modified version of DSR (which they claim as an IETF protocol, though it isn't) that they call Link Quality Source Routing (LQSR).

Their MCL driver implements an interposition layer between layer 2 (the link layer) and layer 3 (the network layer). To higher layer software, MCL appears to be just another Ethernet link, albeit a virtual link. To lower layer software, MCL appears to be just another protocol running over the physical link. It has several other interesting features as well, including support for a proposed improvement to ETX that they call ETT (Expected Transmission Time).

More interesting, given yesterday's discussion, is that Microsoft has already published results on using multiple radios.

Victor Bahl has also published a proposal to allow co-ordinated channel hopping called Slotted Seeded Channel Hopping (SSCH) that requires no leader election or other centralization. This isn't 802.11s Frequency Hopping, but rather a method to allow multiple 802.11b, 802.11g or 802.11a radios to change channels in a co-ordinated manner in order to increase capacity. The paper shows that this can acheive results comparible to using several radios, while the cost and complexity of the resulting solution will be much lower.

There are other interesting papers on the page as well. here they show that ETX is a near-optimal routing metric when all the nodes are stationary (as they would be in a neighborhood mesh network). Presumably they then discovered an improvement (ETT, above).

There is even a paper co-authored by Jim Kujiya on building a low-cost steerable array using dielectric phase shifters. I met Jim Kujiya back in 1991-1993 when Jamie was acting as the Electronic Theatre chair for SIGGRAPH 93. Kujiya is probably best known for his work on the Evans & Sutherland frame buffer, so its good to see him doing things outside of computer graphics.

So if you're a VC who has funded a mesh-based product company, you might want to ask a few questions at your next board meeting. Questions such as, "Why didn't you tell me that Microsoft is doing mesh?" and "Why did you (use my money to) pay to license TBFPR?" And then ask yourself how you're going to make your investment back from a company who's only competence seems to be in using other people's hardware with other people's software?