GigaOm rounds up the low subscriber numbers at early metro-scale Wi-Fi networks: They note the SF Chronicle story from yesterday about Taipei's underrun on necessary subscribers; point to Lompoc's 281 subscribers on a $3m network, where 4,000 are needed to break even (based on a 2003 analysis); and to criticism of MetroFi's first-phase Portland, Ore., rollout. EarthLink wouldn't give Katie Fehrenbacher usage numbers for their early networks.
Fehrenbacher concludes that mobile workforce applications will drive metro-scale networks--as Craig Settles noted--and that the way these early networks have been deployed, they don't provide either high speeds outdoors nor good indoor coverage.
As I've been writing for some months now, it's been very clear that the expectations set have been too high based on network density. It seems clear that to achieve indoor access rates at the level that residents expect, not the level at which a network can be financially and technically optimized, may be far above the density of nodes that are being built in most networks.
Of course, these are still early networks. Taipei is the largest single deployed network, although nothing like the largest network planned. They made some missteps in rollout, including having no compelling applications--not even a deployed VoIP service--that would change usage patterns. Lompoc has had equipment and vendor problems, and is rejiggering prices. Portland, Ore., is barely deployed, with just a few percentage points of the initial network installed.
St. Cloud, Flor.'s mayor just told me that every antenna in their network has had to be replaced due to water damage. They didn't pay for the antenna replacement, but city workers' time was apparently involved. St. Cloud scored highest in a recent network evaluation of several deployed U.S. Wi-Fi systems, but even they're having teething pains.
Update: The folks at Meraki sent me a link to the live overview of usage in their San Francisco network. As a free network with exposed stats, it's pretty interesting to see what usage is like on an ongoing basis.
Glenn, as you know, I don't rep Cohda Wireless anymore, but this is a good informational site on why things are melting down in these cities (i.e. limitations of outdoor Wi-Fi).
www.thedirtylittlesecret.net
Here's a bit more information on the St. Cloud antenna situation. It came to Tropos attention a while back that a manufacturing defect, subsequently corrected, caused us to receive from our antenna vendor a batch of antennas that could develop problems with water ingress. We notified our customers who were potentially affected and worked with them to test their routers to see if any of their antennas might have problems. A small percentage of the antennas in St. Cloud were suspect. The city chose to be cautious and replace all of their antennas. The antenna replacement is now complete.
Glenn, you are getting close to what the real problems are with these networks.
What needs to be detailed is the vendor Mesh products being used in these networks and then we will be able to see the trend evolve.
I think the Taipei network uses the Nortel Mesh (Dual Radio)and Earthlink deploys the Tropos (Single Radio) devices and MetroFi uses the Skypilot (dual radio).
These problems need to be documented and highlighted quickly in the analysts community so the industry (Wireless Metro Area Mesh Networks) are not given a bad name by poor decisions on the part of Consultants and Muni's in selecting the right technology (Mesh Nodes). Seems to me these folks are picking a Provider (ie Earthlink or MetroFi) vs the righ technology. The consultants should know better and need to consider both.
Big concern:
The Cell Carriers (and new WiMAX providers)will quickly take advantage of these high profile poorly designed/deployed Mesh Networks by positioning their Narrowband (5-700Kbps) networks as a better solution for commercial customers (big revenue source)and the Muni's themselves.
I have yet to see or hear of a Metro Area Mesh Network deployed by a competent Local Service Provider who knows what a Carrier Grade deployment is and has made sure that the network they deploy can be upgraded (adding 802.11n, WiMAX and or the new 4.9Ghz systems in the field) and scales without having to add more nodes per square mile. I actually heard one vendor (big router company) say that a good network is one that has the most Nodes per square mile-what amazing marketing spin.
Poor support: What I think is really happening here is users of these networks log in and find these networks wanting (limited coverage/reach and minimum bandwidth)and elect to avoid them. What do you think is going to happen to these networks when the users begin to use them for VoiceIP, Gaming (P2P) and video and photo sharing ?
Customers want these portable network and will pay (premium) to allow them to take their Broadband Wired service (DSL/Cable Modem)with them when they go portable.
Sorry, frustrated by the obvious.
Jim
I just checked the meraki SF map you linked to. Almost all the nodes in the Map are gateway nodes (i.e. they get their Internet connection from a standard wired internet source like DSL, Cable Modem or whatever).
There are a few nodes that are 1 hop from a gateway, all serving one or two users. A couple had 2 hops.
This is not a map of a mesh network!
The dirty little secret isn't a problem in PHYs. The dirty big secret is that today's 802.11 MAC will not support the kind of scale we need for mesh networking.
The distance between APs is not determined by how strong a signal an AP can emit, but by how strong a signal the clients can emit. One of the biggest problems I've seen is that clients can hear the APs fine but the AP can not hear the client. But in many cases the client's software thinks that if it can hear the AP fine, then everything is fine and in many cases will use high modulation rates to talk back to the AP making the situation even worse. And the UI of the client will even say it has a great connection but then not be able to communicate.
So the solution to this is high density deployments. But if you do high density deployments, then the APs will all be contending with each other at the MAC layer. It will work very nicely when there is little or no load (ie no users, like when giving a demo). But as soon as there is any load, the traffic between all the nearby APs will start to fill with retrys leading to collapse of the medium (those of you old enough will remember the same situation when we had Ethernet without switches, just hubs and bridges)
This will be a problem until we get an improved MAC. I'm hoping that 802.11s will make a contribution to solving this problem.
Note that single radio mesh like Tropos and Mereki will suffer the worse from this issue as the mesh backhaul is in the same contention domain as the client access.
Two radio mesh APs will do a bit better since they isolate the client traffic from the mesh backhaul, but if you are using vanilla 802.11 mac for the backhaul, the APs will self-contend. Skypilot uses a modified, non CSMA MAC for their mesh so they have this problem to a lesser degree at least for adjacent nodes. But they will still have the problem for their client radios that are near by.
I believe the MAC issue will eventually be solved, just as it was in Ethernet. But in the mean time expect continued negative hype cycle to reach at least as high as the positive hype cycle was!
[Editor's note: Robert is extremely credible, but has consulted for SkyPilot, which should be noted. This doesn't make him any less credible to me.-gf]
I really hope EarthLink survives, but whats going to happen to the cities they are in if they fail? Some cities are loading up a bunch of city services on the network, but what happens to the network if the private company who owns it is not turning a profit? Will the city as the sole anchor tenant be able to generate enough profit for EarthLink to maintain and update these networks?