Thought the Herald survey was a bit meh, so started work on one myself. It's pretty basic, but its functional. Hoping to get some people to try it out submit their results to populate the database. At this stage it just compares national and international speeds, but has the potential for more.

http://bgrade.bage.net.nz/

Code Average speed (int dn): 32.61KB/s Average speed (nat dn): 439.21KB/s Average speed (nat up): 16.43KB/s

The client is written in PHP (not an ideal language, though it does work on Linux!) , and i've included a compiled .exe for people without PHP installed.

RSVP with suggestions - ie. what would make people use this as opposed to speedtest.net (don't say make it flash based!).

Threarah wrote: what would make people use this as opposed to speedtest.net (don't say make it flash based!).

If you're going to try and turn it into an accurate database and a more public thing than just a proof of concept you're going to have to have an easy to use web based GUI, with no executables. AFAIK flash is the best way to accomplish this.

Average speed (int dn): 198.85KB/s
Average speed (nat dn): 645.9KB/s
Average speed (nat up): 53.36KB/s

Test timestamp: 14:26 19/11/2008
IP address: 203.100.223.xxx
ISP: Unknown

Time of tests: Tue Nov 18 22:54:29 2008
Average speed (int dn): 36.46kbps
Average speed (nat dn): 293.85kbps
Average speed (nat up): 38.39kbps

Threarah wrote: RSVP with suggestions - ie. what would make people use this as opposed to speedtest.net (don't say make it flash based!).

Great work btw!
If I was to suggest anything I'd say have it web based, with a bunch of summary stats about the different ISPs that have taken the tests. % or ratio results might work well as then it wont matter about the actual speed and line sync rates., eg, the test shows that the ISPs is providing say 20% international performance compared to national performance.

All plotted to a 24hour graph for easy reading. Soon the results will start to add up and a nice graph will appear similar to the isposure/epitiro report but with % or ratio results and not simply raw speeds only.


Time of tests: Tue Nov 18 23:59:13 2008
Average speed (int dn): 192.35kbps
Average speed (nat dn): 734.64kbps
Average speed (nat up): 54.1kbps


sure thats not KB/s and not kbps?
Maxnet

Time of tests: Tue Nov 18 23:57:50 2008
Average speed (int dn): 25.13kbps
Average speed (nat dn): 521.05kbps
Average speed (nat up): 41.29kbps

Average speed (int dn): 179.7KB/s
Average speed (nat dn): 501.93KB/s
Average speed (nat up): 55.07KB/s

Maybe just link direct to the exe so noobs can figure out how to run it and have it load a page with your specific results once it's completed?

Looks like you'll get some nice data here

Would be much more 'noob' friendly if it was hosted with a GUi, where all you do is press one button, it does the test and then loads the results page.

Could you host it in a ZIP or just as the EXE itself? Can't open RAR here!

Xnet coming in last place!!

Average speed (int dn): 39.33KB/s
Average speed (nat dn): 446.08KB/s
Average speed (nat up): 14.65KB/s

Go Large ftw

Although it's mainly the YouTube test limiting people's internatonal results, Google probably have QoS on it so won't really go much faster than 200KB/sec


Steam and YouTube averaged it down quite a lot, throughput during st server tests were similar.

Good job Ben!

Code Results for test #66 Average speed (int dn): 189.62KB/s Average speed (nat dn): 289.01KB/s Average speed (nat up): 57.15KB/s Test timestamp: 10:24 20/11/2008 IP address: 222.152.94.xxx ISP: Xtra/Telecom

I dispute that national speed test, It should normally be much faster than that!

Done a bit of jimmying:

Can download a .zip or .exe
When its done, it now opens this page - means I can add more result info later on without having to update the client
Made a more relevant result page - takes all data from peak (3pm-12am can easily change though) times, gets rid of those that might have shoddier lines (national speed has to be over 200KB/s) then sorts by int speed

Do all go-large addresses resolve into *.broadband-telecom.global-gateway.net.nz? Might separate them from the other Xtra/Telecom results if they do.

And *.wxnz.net addresses are xnet yeah?

Threarah wrote: Made a more relevant result page - takes all data from peak (3pm-12am can easily change though) times, gets rid of those that might have shoddier lines (national speed has to be over 200KB/s) then sorts by int speed

This may not work as you expect and may result in taking out results that are genuinely shoddy

Threarah wrote: Do all go-large addresses resolve into *.broadband-telecom.global-gateway.net.nz? Might separate them from the other Xtra/Telecom results if they do.

Yes.

They have moved the ranges around a little in the past months but generally
125.238.0.0/16
125.236.128.0/18 (used to be /17)

Mental69 wrote: This may not work as you expect and may result in taking out results that are genuinely shoddy

True that. Next job I guess would be to make custom result filters, so people can sort by their own criteria.

for example me running it from work will throw it off.
Should maybe work on some ISP averages or something.

Malks wrote: for example me running it from work will throw it off. Should maybe work on some ISP averages or something.

The plan is to graph the average speed against the time of day (possibly weekday/end too) for each ISP.

Code Average speed (int dn): 68.34KB/s Average speed (nat dn): 450.64KB/s Average speed (nat up): 16.27KB/s Test timestamp: 13:00 1/1/1970 IP address: 123.255.31.xxx ISP: Snap (if this is wrong let me know)

TARDIScode!

Threarah wrote: Done a bit of jimmying: Can download a .zip or .exe When its done, it now opens this page - means I can add more result info later on without having to update the client Made a more relevant result page - takes all data from peak (3pm-12am can easily change though) times, gets rid of those that might have shoddier lines (national speed has to be over 200KB/s) then sorts by int speed Do all go-large addresses resolve into *.broadband-telecom.global-gateway.net.nz? Might separate them from the other Xtra/Telecom results if they do. And *.wxnz.net addresses are xnet yeah?

Yay, that's exactly the implementation I had in mind too

Could you delay opening the URL by another 10 seconds or so? It loaded the website before the data was recorded for me, I had to refresh.

With the national tests your only really testing who got the fastest, uncongested connection to APE. Which is a bit unrealistic as for most peopl what you really want to know is how good x ISP's connectivity is to Telecom and Tesltra as unfortunately most content is still on their networks and they don't peer at the exchanges. Maybe you should add a test to something hosted at telecom and something hosted at Telstra

wjw wrote: With the national tests your only really testing who got the fastest, uncongested connection to APE. Which is a bit unrealistic as for most peopl what you really want to know is how good x ISP's connectivity is to Telecom and Tesltra as unfortunately most content is still on their networks and they don't peer at the exchanges. Maybe you should add a test to something hosted at telecom and something hosted at Telstra

I can tailor the national speedtest to whatever ISP the user is from (could dynamically generate the sources page) - doing that could give each ISP a national test on their own network. I'm trying to figure out graphing in PHP atm.

Shade wrote: Could you delay opening the URL by another 10 seconds or so? It loaded the website before the data was recorded for me, I had to refresh.

Eek! It shouldn't open the result page until it gets a response from submitting the results - and it'll only get the response once the sql stuff has been run.

Anyone care to suggest a name for this project :P?

Got some rather budget ISP average graphs going - no titles or legends or anything, but still