The metropolitan mesh equipment vendor Tropos continues its juggernaut-like movement, releasing more ways to bind others into their fold: The company's announcements today are about reaching out to other firms. They announced a software API (application programmer's interface) that allows third-party developers to extend their own software or management interfaces to work down to the level of Tropos equipment. The partner program provides these developers access, and the list of partners is quite extensive.
They also previewed three tools for providing better operator pre-build planning, operation, and monitoring: Insight, Drive, and SignalPro by EDX Wireless with Tropos modules.
Although the press release lists them in reverse order, let's start with SignalPro: the Tropos modules will allow metropolitan-scale radio frequency planning using topographical and other data. It's the software used for planning cellular and other wireless networks today.
Tropos Drive lets an operator test deployments by emulating different hardware clients and correlating that with GPS coordinates while driving. A neighborhood's effective coverage can be measured and changes can be made before end-users are involved.
Tropos Insight "identifies places where the network throughput and capacity can be improved by looking at the backhaul links, the intramesh links, and also the client performance," said vice president of marketing Ellen Kirk said.
The range of what Tropos expects to see its partners integrated is across the board with access, public safety, and government functions. "Everything from video surveillance, municipal automation, indoor equipment to feed the delivery of these solutions: it's not just about offering connectivity any more from Tropos," said Kirk.
The recent addition of Kirk to the Tropos management team reveals its direction: She comes from Qualcomm via SnapTrack, and was at AirTouch before that. Cellular industry know-how might come in handy in metro-scale networking.
The array of partners signing on to the Tropos program is extensive, but Ruckus made a, well, ruckus with their own press release today: they're producing the MetroFlex, a CPE (customer premises equipment) device using beam-forming technology designed for what will probably be a multi-million-unit per year business by 2007. This isn't the full on, multi-stream MIMO, but it is a range-extending technology.
EarthLink, for one, expects that its retail network partners in various cities will need to provide CPEs to customers so that the signal is strong enough within their homes and businesses. (This isn't new, although some anti-muni types have pretended it was. The CPE requirement means that 802.1X authentication can be pre-configured before being shipped to an end user; EarthLink will require this strong authentication for login and network encryption for its networks.)
Because EarthLink has apparently standardized on Tropos as their mesh and end-user connection equipment, the Ruckus CPE will likely be an early choice for the Internet service provider to recommend. With the Tropos API and partner program, a service provider will be able to monitor and respond to problems all the way to the customer edge in the CPE in their home, rather than ending at the access point.
Tropos's MetroMesh Solution Partners, as they term them, includes Motorola (Canopy will be the backhaul in EarthLink networks), a few smart meter firms, and a host of previously partnered OSS/clearinghouse companies that handle back-end user authentication, billing, and roaming, including Boingo Wireless, Airpath Wireless, and Pronto Networks. The list also includes NetMotion Wireless, a firm that uses client and server software to allow seamless transitions across any network medium for enterprise customers--and, I expect, government networks as well. They have a killer demo.
These guys have so much momentum in the marketplace it's just silly. Juggernaut is right. And it's not just because we're a "partner." We're everyone's partner. Tropos' routing code is simply spectacular and they have more hands on experience in this market than everyone else (including Cisco who has sold a lot of wireless bridges....eeew weee....combined). By the time WiMax gets here, quite frankly, there might just not be a market for it with Wi-Fi metro networks everywhere. Then again, what the hell do I know?
Tropos has brilliant marketing with a fabulous CEO who is everywhere and has optimized marketing. They now have huge momentum. Too bad their product architecture is severely flawed.
It will be very interesting to see what happens when the Troppos networks have to actually carry some load.
Routing is not the hard problem of mesh networks based on 802.11. The combined requirements of density (to deal with the fact that 2.4Ghz does not go thru trees or buildings) and deliverable bandwidth (ie to get broadband speeds) means that 802.11 MAC contention is the big problem.
This is the same problem that the original Ethernet (remember the single cable or hub Ethernet? Most of you are too young so you will repeat history) Just like the original Ethernet, standard 802.11a/b/g/n will work great with little or no load (ie only a small number of simultaneous users).
But as soon as you start adding simultaneous access to 802.11 APs, contention becomes the bottleneck. And it is not a linear degredation, its an explosive degredation. After about 8 simultaneous users (and I mean truely simultaneous access of signals, not necessarily just 8 end users websurfing, but maybe 8 clients simultaneously doing heavy FTPs or worse yet, streaming video) the AP thruput will start degrading badly. A few more than 8 and it will collapse and there will be nearly no thruput unitl some of the clients stop using it.
This is because the 802.11 MAC protocol is a collision sense multiple access medium. There is no mechanism (in most 802.11 implementations) to allocate capacity in a controlled manor. Nodes communicating via 802.11 must first listen before transmitting. If they hear another node transmitting, they must shut up for a period of time. If they do transmitt and then discover that they collided with another node that had simultaneously transmitted, then they both have to back off. If they collide again, they have to back off an even longer tiime.
So you can see that if there are not many users, than each user gets almost all the capacity of the channel. (and thus 802.11 can give great demos!) But if you get a number of simultaneous users, the channel capacity gets full first by the use, and then explodes with retrys. So the retrys overwhelm actual thruput. Thus the channel collapses.
Tropos is the worst of all worlds in terms of contention (though there are many other vendors that have a similar problem). Tropos uses the same single 802.11b/g radio for connecting clients to the mesh and to relay to other mesh nodes.
So not only do you have the local contention of client nodes, but every client access will generate a mesh node transmission, immediately doubling the number of contending transmissions.
Troppos does have one of the best 802.11b/g radios. They really do support 1 watt with very nice band edges. But all that does in increase the contention domain. Many more nodes in a neighborhood mesh will hear each other, causing even more contention THROUGHOUT the neighborhood mesh. So the very high powered radio is actually a problem when used in dense municipal wireless environments. (And it will still not penatrate buildings or many more trees).
Now if Tropos deployments are not really using Tropos to mesh, but are really using the Troppos units as super APs and are really using Motorola Canopy's for all the "meshing" (really relaying, since Canopy's are point to multipoint) then maybe you only have to worry about local access contention. But even there the high-powered APs will self interfere for the access portion, but then Tropos shouldn't be considered a mesh solution.
These problems will not show up to much until there is some real load on the network. I will be very interested to hear when people do have such loaded networks and find out if what I predict is true or not.