MetroFi’s Portland network slipping into unmitigated decay?
Posted by Russell Senior on 5 January 2008A couple weeks ago we commented on how, despite earlier assurances, dozens of new access points did not appear to be going live over the last few months. We had based this observation on 
a comparison of roughly weekly snapshots of the MetroFi map. Since October 26, 2007 the map has recorded the following changes:
- between October 31 and November 2, a “Coming soon” access point appeared on the map at 4400 NE Broadway
- between November 9 and 18, that “Coming soon” access point disappeared from 4400 NE Broadway and reappeared at 416 NE Brazee
- between December 1 and December 14, the “Coming soon” access point at 416 NE Brazee was replaced with an “In service” access point at 2432 NE Martin Luther King Jr Blvd. In the same period, a latitude/longitude correction was made, moving the spot on the map intended for SE Division Street and 57th Avenue about 1000 feet from SE Lincoln to SE Division.
When Chuck Haas said that the service area had expanded substantially since October, we were perplexed by the disparity and inquired to MetroFi about just which new access points had gone live since October. We are still waiting for an answer. Their map indicates anemic growth at best. But then we began wondering, just how accurate is that map? There is only one reliable way to find out. Go check. So, between December 19th and Decemeber 30th we spent about 35 hours doing just that. From the locations provided on their map, we endeavored to drive underneath each one of the 677 SkyPilot access points that MetroFi claims exist.
Using a wifi network mapping apparatus with a wifi antenna and GPS device mounted on the roof of a car, we drove to the locations indicated by MetroFi. If the SkyPilot was not where MetroFi said it was, we checked on neighboring streets and sometimes used radio signal to zero in on it. When we visually located the SkyPilot, we drove as close as possible, usually directly under it. Since each SkyPilot emits a unique 6-byte identifier (called the BSSID) in its “beacons”, by driving directly under the SkyPilot we would be able to link each SkyPilot on the map with a high signal strength measurement from a particular BSSID.
The wifi network mapper distinguished three conditions:
- Beacons containing a network name, or SSID, of “MetroFi-Free”;
- Beacons containing an empty SSID; or
- No beacons at all
Results:
- The MetroFi map claims 677 SkyPilot access points:
- 588 are indicated as “In service”
- 88 are indicated as “Coming soon”
- 1 is not indicated as either (at SW 1st and Meade).
- Beacons were heard from 656 SkyPilots during the survey.
- One of the SkyPilots was heard only very weakly, near the crest of the West Hills. It is not clear where that SkyPilot might be located.
- All of the other 655 beaconing SkyPilots were located:
- Four of these SkyPilots do not match any of the locations on the MetroFi map.
- SE 92nd and Hawthorne Blvd
- SE 90th and Stark
- SE 76th and Stark
- SW Patton Road and Hillside Drive
- Of the remaining 651 located SkyPilots with beacons:
- The median distance between the indicated map location and the location of peak signal strength overall was about 28.5 meters (this distance probably represents a reasonably good map location);
- For 33 SkyPilots, the peak signal was found over 100 meters from the indicated map location.
- For 9 SkyPilots, this distance was greater than 200 meters.
- For 5 SkyPilots, this distance was greater than 400 meters.
- The greatest discrepancy in distance observed was about 2400 meters.
- Four of these SkyPilots do not match any of the locations on the MetroFi map.
- 23 SkyPilots were located but for whatever reason were not emitting beacons. Of these, 17 were marked “In service” during the period of this survey.
- 3 SkyPilots were not found at the designated location and no unidentified beacons were heard nearby.
- N. River and Essex
- SE 41st and Liebe
- SE 9th and Division
- In total, 678 SkyPilots were located (beaconing and non-beaconing), one more than were listed on MetroFi’s map.
- One of the SkyPilots that MetroFi tags as “Coming soon” (near SE 34th and Clinton) was emitting “MetroFi-Free” beacons.
- Of the 588 access points that MetroFi claims were “In service”:
- 1 could not be located (SE 9th and Division);
- 17 were not emitting beacons (and thus were not in service);
- 18 were emitting empty SSIDs (and thus were also not in service);
- 552 were emitting “MetroFi-Free” beacons for at least part of the period of this survey.
Conclusions:
- Generally, the quality of location data on the MetroFi map is disappointing, particularly when you consider that each SkyPilot contains a GPS device and publishes its location information in multiple formats readily available to MetroFi. Any SkyPilot with which MetroFi has a network link should be able to tell them precisely where it is located. The large errors found in some of the access point locations is therefore hard to fathom.
- The quality of the access point status information on the MetroFi map is also disappointing. At least 36 of 588 “In service” access points were found to be inoperable from a potential users perspective. This information ought to be readily available to MetroFi.
- Assuming all of the SkyPilots found to be emitting MetroFi-Free beacons were actually capable of providing an Internet connection (there is some reason to doubt this), and assuming (as we found in our earlier Proof-of-Concept survey) that the SkyPilot devices only provided a 90% probability of a connection in outdoor areas (the standard agreed to in the contract with the City of Portland) within about 300 feet of an access point, and assuming none of these coverage circles overlap, then the total fraction of 134 square mile Portland being provided coverage during this survey is approximately 4.2%, not the 29% claimed by Chuck Haas.
- There appear to be at least 125 SkyPilots hanging on street lights in Portland, a retail value of about $225,000, that apparently aren’t providing any service to Portlanders or generating any revenue for MetroFi.
- The significant number of non-functional access points labeled “In service” gives one an impression not of steady-if-slow growth, but of neglect and decay.
One Comment
Leave a Comment
Read more
« MetroFi rollout not making visible progress
Slight modifications to MetroFi’s map appear »
January 7th, 2008 at 5:09 pm
You guys are awesome! Thanks for continuing to do this work.