This might be useful to some.
Article :
http://www.reuters.com/article/idUSTRE62B08720100312
site :
It requires giving your address.
Correction: it _requires_ Java. It _asks_ for your address. It seems
like it'd work fine if you gave it your neighbor's address. 
I noted that I got wildly varying numbers on a laptop and an iPhone (there
is also an iPhone app) and the iPhone app doesn't ask for an address. Both
on the same wifi, and the numbers were off by a lot.
... JG
Joe Greco wrote:
Correction: it _requires_ Java. It _asks_ for your address. It seems
like it'd work fine if you gave it your neighbor's address.I noted that I got wildly varying numbers on a laptop and an iPhone (there
is also an iPhone app) and the iPhone app doesn't ask for an address. Both
on the same wifi, and the numbers were off by a lot.... JG
INSTEAD of using the FCC provided app, one 'could' always use OOKLA and
M-LAB directly.
The following links may prove to be more helpful to some.
http://demo.ookla.com/linequality/ *and *
http://npad.iupui.lax01.measurement-lab.org:8000/ (Choose the closest
orig/term point to you from:
Measurement Lab Tools - M-Lab )
Both sites present varying granularity.. It goes without saying that
one should NOT send one's mother/grandmother to the NPAD site. Pete
(Peter L�thberg) being the exception here..... O:-)
Best,
Robert.
So have other people noticed that the Ookla/Speedtest.net/Speakeasy
Bandwidth test often comes up VERY short on upload bandwidth results for
anything other than residential-grade asymmetrical services?
We often get complaints from customers saying "I'm not getting the upload
bandwidth I'm paying for", and when we ask what they are using to determine
this, the answer is almost always either Speakeasy or Speedtest.net.
We certainly don't depend on or recommend these sites to customers (we have
our own internal tools and usually recommend FTP or iperf), but everyone who
deems themselves semi-knowledgeable seems to find their way there anyway.
Do these sites simply not have the downstream bandwidth to handle the upload
tests? If thatÂ’s the case I'd really like to see the admins add a
disclaimer of some form directly to the site.
  Thanks,
-Scott
Scott Berkman wrote:
So have other people noticed that the Ookla/Speedtest.net/Speakeasy
Bandwidth test often comes up VERY short on upload bandwidth results for
anything other than residential-grade asymmetrical services?
The question to consider are: are JAVA based "speed" testers reliable?
What are the caveats?
Nevertheless, they have - over time, become 'popular' among users, and
your average 'cable guy,' to at
least the first 2 tiers of service tech personnel at ISPs who are often
relegated to diagnosing why "grandma's"
Internet connection is slow....
Best,
Robert.
I'm seeing big disparity between upload and download speeds. I had the
same thought as to the testing platform expecting asymmetrical speeds
typical of a residential link.
Why didn't they go with NDT?
We decided to spend the money to install a local Ookla speed test
site a couple of years ago and have been happy with the decision:
1) Local customers who run the speed test get much more accurate readings
than with what we were previously using, which was either javascript based,
or java based. The Ookla software we're running is flash based, which a
very high number of our users already have installed.
2) It gets placed on the main speedtest.net map. When our customers go to
speedtest.net to test their speeds, the default test location they get is
our own site, and they get accurate results.
3) When customers from local competitors go to speedtest.net, they get
defaulted to our test location, and get less than accurate readings (since
they are not on our local network) and get artificially depressed results,
which is a positive for us.
On a side note, we've tried to sign up for Ookla's pingtest.net, but
haven't gotten any responses from them about it. Has anyone else had any
success signing up for it?
There is definitely something very broken in the gov't version of the
speedtest.net application. It seems very BW constrained. I can get great
results to a variety of ookla sites via test points across the US, but the
government one is always horrible.
We host both a pingtest and speedtest.net site, and provided you give it
enough BW, it's reasonably accurate for connections up to around 50 Mbps..
Peter Kranz