Questions for Tropos: Does Google's Mountain View Network Fold Under Pressure?

This was recently sent to me and raises some serious questions about what's really happening on the Google/Tropos Mountain View Network when load levels increase:

Subject: Re: SF Wifi: How good is the coverage for Google's Mountain View network?

This is typical with Tropos radios.

One of my closest friends is a contractor for FEMA and occupied an apartment in French Quarter of New Orleans. His balcony was located across the street from a node. He told me even sitting on the balcony and trying to check email, it was giving him about 28k best effort.

It's amazing to see what government, excuse me, taxpayers are paying for. The economics of the system deployment and monthly reoccurring is unbelievable.

The system described below uses single channel/frequency radios.

1. This architecture restricts load capacity on the systems ability to aggregate TX/RX traffic not just how much band-width is available. Once stressed, the system errors begin to compound at a rapid rate and ultimately crash.

2. The amount of nodes required to cover a square mile is about 25, based on the OEMs calculations. This limits the root to repeater ratio to 6:1. Meaning that for every 6 repeater nodes there has to be a gateway or root node. The proper way to interface the wireless to the core WAN is at each root with a terrestrial connection, i.e. T-1 etc. Most are manipulating the system by using another layer of back haul infrastructure such as a point to point system using an alternate frequency. This will get you right back where you started. Refer to #1 comments about issues with this architecture.

3. If a WAN interface is structured to each root node than you have on average 5 WAN interfaces per square mile. Let's say that a T-1 connection will cost you $250 per, you do the math. This is before over head, maintenance, SLA burn etc. This formula is also for best possible conditions and wouldn't take into account variables such as, buildings, noise, interference, obstructions, and elevation of node location.

So it seems that marketing and hype seem to make the decisions not the past or current performance of an existing system, it just can't preform when loaded.

Some users on the Google Message Board had this to say concerning coverage:

• "My house is on California St. (cross street is Rengstorff) and I can rarely have the Google WiFi stable connected by using Ruckus MetroFlex device (the PepLink is worse). I cannot use Google's network even though house (and window) is by the street, so I cannot imagine how the people who are living away from the street can access Google's wifi network." (Posted by "weili")
• "I live near Farley St. and get an 'OK' connection, using my HP V2410's Broadcom WiFi adapter. I work near Shoreline and get about the same speed. I once tried connecting near Therkauf (sp) elementary school, and couldnt get connected there, even though I was right underneath an access point!" ("wayshwing")
• "While I can see and connect from inside my home, the service isn't useable at all no matter where I am while inside. Yet, if I go outside and position myself near Google's AP on the street, the signal from my own cheapy AP is strong and very usable." ("Brian")
• "I am about 400 ft from the nearest Google WiFi Node and I had one of the three Google-suggested companies come by and test and set up a system. After two hours of trying both Ruckus and Peplink systems we could not get any connection from my house." ("DJA")
• "I have been trying to acess GoogleWiFi for the last two months with very spotty success. I am on N Shoreline near Middlefield, between two nodes. I use a laptop on the second floor of an apartment complex. I thought the extremely weak and intermittent signal was because of the trees, but I took my laptop outside one evening, past the trees, where I get a clear line of sight to a node (and the other node in the distance). The signal was just as weak." ("uthr")