dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
108115

justin
..needs sleep
Mod
join:1999-05-28
2031

1 recommendation

justin to Dustyn

Mod

to Dustyn

Re: FYI: for general feedback on the new speedtest

ie11 has returned to being really precious after I decreased the waiting around the test was doing. Will add back the malingering.

leibold
MVM
join:2002-07-09
Sunnyvale, CA
Netgear CG3000DCR
ZyXEL P-663HN-51

leibold to justin

MVM

to justin
At the DSL connection it is Konqueror 4.11.5 (correctly identified in the result data), here is another test result: »[DSL Speed test: 5.93/0.00 37 ms]

I'm fairly certain it is actually sending data during the upload test. No idea what is going wrong.
jaberi
join:2010-08-13

jaberi

Member

i am on low speed, so not sure that is helpful to Justin.

leibold
MVM
join:2002-07-09
Sunnyvale, CA

leibold

MVM

If you look at his responses in this thread it sounds like he is interested in getting accurate results for all kinds of Internet connections (fast and slow).

justin
..needs sleep
Mod
join:1999-05-28
2031

justin to jaberi

Mod

to jaberi
This is your result, I'm interested in any speed result if you don't believe it to be about right.
»[Cable Speed test: 15.77/0.41 52 ms]
DrStrangLov
join:2012-03-28

DrStrangLov

Member

RE: Satellite

Ping times are about 70 ms higher than if I ping to same city via DOS prompt.

Just the way it is?
abadaba
join:2005-09-25
Mason City, IA

abadaba to justin

Member

to justin
Not very accurate...Ran many test at all times of the day and downstream still never gets above 35mbps...However pings seem to be consistent...

Fiber 100/100 Time Warner Business Class

»[Fiber Speed test: 24.20/76.51 15 ms]

»www.speedtest.net/my-res ··· 77964064

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin

Mod

said by abadaba:

Not very accurate...Ran many test at all times of the day and downstream still never gets above 35mbps...However pings seem to be consistent...

to eliminate any causes on your side, check with a different browser. I looked at your log and everything looks clean, the servers are not even breathing. so there are two possibilities either your pc has something slowing things down in the browser (but not in flash) or time warner business in your area is only fast to Austin texas but not particularly fast to the net in general.
If you look in the log for your test you can see which locations are contributing. Atlanta is at internap and entirely un stressed and the other locations test google fiber users at 1gig and companies at over 1 gig so they would have no issue filling up your download.
I would like to figure it out with you so send me anything you find by changing browsers or looking at your pc/mac setup?
trix2
join:2012-06-11
Columbus, OH

1 recommendation

trix2 to justin

Member

to justin
Tests still seem accurate here. To be clear, this is operating on IPV4 only?

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

1 recommendation

justin

Mod

yes
I can add ipv6 but since none of the cloud locations support it on the server level, an ipv6 run would have to be to michigan only where I have an ipv6 server. And that probably would run slow because of latency.

(both google and amazon have ipv6 at a load balancer front end level but it screws up test timings).
abadaba
join:2005-09-25
Mason City, IA

abadaba to justin

Member

to justin
Tried with Safari from work...I maintain the links, there is nobody here other than myself and a co-worker.

»[Fiber Speed test: 23.26/76.12 28 ms]

Atlanta speedtest.net

»www.speedtest.net/my-res ··· 78000719

From home 80/20 FTTH GVTC ie

»[Fiber Speed test: 66.83/19.36 44 ms]

My history with the new test from work

Local Time Down Up Ping Streams Comment
2015-04-09 17:20:19 23.26 76.1 28 16 / 6
2015-04-09 16:55:17 24.2 76.5 15 16 / 6
2015-04-09 09:11:30 20.05 61.7 17 16 / 6
2015-04-04 13:30:23 25.5 66.7 20 16
2015-04-02 11:51:01 21.05 43.1 27 16
2015-04-02 11:34:22 20.16 60 27 16
601 MB used in last 30 days
abadaba

abadaba to justin

Member

to justin
From S2012R2 within a virt within ie on our network...There aren't any traffic shapers on the server side and still the same result.

»[Fiber Speed test: 25.65/70.14 41 ms]
trix2
join:2012-06-11
Columbus, OH

trix2 to justin

Member

to justin
Got it. Someday I think it would definitely be worth it. I have been playing around with my settings and it caused this speedtest to fail, which is why I asked. All seems well again though.




The HTML5 based speedtest is truly awesome though.

»[DSL Speed test: 43.08/4.67 54 ms]

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin to abadaba

Mod

to abadaba
can you explain why your work connection goes to 23 but your home link is fine, basically (even with ie11).

do you have chrome at work or firefox (safari is fine though I prefer it on osx).

Its a puzzle. There is something in the way at your office, I think. Unless your work just has one good peering arrangement. What does work do on a speedtest.net to a california or east coast server?
abadaba
join:2005-09-25
Mason City, IA

1 edit

abadaba

Member

SpeedTest to Cali

»www.speedtest.net/my-res ··· 78024896

OldSChool flash from DSLR 65mbps

justin
..needs sleep
Mod
join:1999-05-28
2031

justin

Mod

there is definitely something up with your work setup. let me know if you run work with a different browser or a different pc..
abadaba
join:2005-09-25
Mason City, IA

abadaba to justin

Member

to justin
If there was something weird with my work setup the speedtest.net test would show the same...I've used Safari on Speedtest as well...To Cali/Tx/Florida etc...all over 80Mbps on virtuals/laptop/desktop

MarkRH
Premium Member
join:2005-02-08
Edmond, OK
ARRIS BGW210-700
ARRIS TM3402
Asus RT-AC68

MarkRH to justin

Premium Member

to justin
Still seems ok: »[Cable Speed test: 118.81/10.40 17 ms] and one from Speedest.net: »www.speedtest.net/my-res ··· 78060442 (117.43/12.00).

Firefox 37.0.1 in Win 7 Pro 64.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin to abadaba

Mod

to abadaba
I'm not saying you have a work problem, I'm just searching for the reason that the html5 speed test isn't working on your work connection when that same work connection can do 60mbit at your home, 1gig on fiber, and so on and so forth.

Usually these things come down to an extension problem, some security software, or something like that.

Clearly whatever it is isn't causing a problem for flash (even our old flash speed test with one stream), hence I am very interested in knowing what it is. I can't debug it for you but suggest that if you explore the issue using a different browser, or different PC (but the same speed test running the same way) you might get some evidence that leads us to the solution and that might be helpful for other people who strike the same problem.
xmd
join:2009-06-25
Carlsbad, CA

xmd to justin

Member

to justin
BROKEN!!!
in a bad way.

Just ran the test on OS X 10.9.5 with Safari 7.1.5 : the test starts running, pulls 100% of my bandwidth (95.7mbps) and hangs up - it's been going for over 90 seconds now: Safari Web content is stuck at 72% CPU, and Safari Networking is stuck at 204% CPU. AND my download bandwidth is fully saturated.

Not a good thing to have a speediest which can fail in a "Use ALL CPU and ALL bandwidth".... Fix it!

[Update]: I went back to the tab in the browser, and the Cancel button did work, but still - nasty failure mode.

If this helps for debugging: I do have a dual WAN setup with a Peplink load balancing router. Happy to do more debugging if I can.

justin
..needs sleep
Mod
join:1999-05-28
2031

justin

Mod

change number of streams to 8, in preferences, does it work ok then?

Edit: I added detection for this situation - running without reporting, for too long. It should error out with a message about one or more streams failing to start.
xmd
join:2009-06-25
Carlsbad, CA

xmd

Member

I set both upload and download streams to 8, and re-ran the test. Works much better this time, although I feel like the upload speed calculation is wrong. My upload bandwidth is about 10mbps total. Check out »[Cable Speed test: 94.84/3.03 30 ms] which shows it at 3.0mbps. In addition, when I ran the test the graph clearly showed that it was above 5.0mpbs for the vast majority of the run - only dropping off in the final second. An integral of the curve clearly would be above 5mps - but the calc says 3. does not compute.
xmd

xmd

Member

I re-ran the test with 2 up / 2 down streams, and this time the download was 85 (which is below the 95 it was actually doing at peak) and the upload shows as 8.5 : closer to my 9.9 upload but still inaccurate compared to what speedtest.net is showing me...
»[Cable Speed test: 84.91/8.46 30 ms]

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

1 recommendation

justin

Mod

Upload on cable is very much influenced by use in your area. The upload of 3.0 was accurate for the duration of the test. Because it peaked at whatever doesn't mean that should be the report if it stalled during the latter part. It ended at 3.0 because that was the overall throughput for the duration of the test. Reporting the peak amount would be misleading.

On a solid connection the peak is the same as the total throughput, but frequently this is not the case on a cable connection where there is a lot more competition for upload bandwidth.

If you have a network monitor take a look at the shape of the upload data. It would be high, for part of the upload - perhaps 9.9 that you think you have, but it would not maintain that, hence the lower report.

Actually I'd say that 8.46 is basically 9.9, within the bounds of variable usage etc etc. But I encourage you to watch the bandwidth during an upload phase to see the difference between peak and actual throughput.
xmd
join:2009-06-25
Carlsbad, CA

1 recommendation

xmd

Member

"Upload on cable is very much influenced by use in your area. The upload of 3.0 was accurate for the duration of the test. Because it peaked at whatever doesn't mean that should be the report if it stalled during the latter part. It ended at 3.0 because that was the overall throughput for the duration of the test. Reporting the peak amount would be misleading."

I disagree: I didn't save the graph, but 99% of the graph was ABOVE 5mpbs (only dropping off at the very end of the test). Simple math says that if 99% of the speed was above 5mbps, there's no way the average can be below 3mpbs, no matter how slow the last 1% was. (Well technically, if the last 1% were *below zero speed* then I suppose it could average out, but that would be a calculation bug).

I suppose it's possible that the calculations are right but the graph is wrong, but I doubt it.

I'll run some more tests and see if I can reproduce it and save the graph too.
xmd

xmd

Member

Ok, it's easy to reproduce : i just ran it again (with 8 up / 8 down) and the average upload is reported as 4.38mpbs but the graph clearly shows that the entire time the speed was between 5mpbs and 10mpbs. (I do note, however, that the graph has two colors: a dark red line and a faint pink line, which is much lower. Perhaps the graph doesn't mean what I think?)

»[Cable Speed test: 87.74/4.38 31 ms]

But in any case, my line is rock-solid reliable at about 95/9 (by several other speed tests) and so far this new one you are offering is all over the map. Until convinced otherwise, the logic says that the new one is not working.

justin
..needs sleep
Mod
join:1999-05-28
2031
Billion BiPAC 7800N
Apple AirPort Extreme (2011)

justin

Mod

the graph is not instantaneous it is speed to date (for upload). So if it drops at the end it is because instant speed actually falls much faster and much further.

the instant speed is actually the faint bars behind the upload graph, once per second.
justin

justin to xmd

Mod

to xmd
of course feel free to experiment with chrome or firefox? every browser instruments upload progress differently, but the ultimate proof is to watch the ethernet interface and check its instantaneous speed. Your results when low show the upload instant speed gets very slow for a few seconds and that drags down the result. To explore whether the browser is confused try chrome?
xmd
join:2009-06-25
Carlsbad, CA

xmd to justin

Member

to justin
Ok, if the graph is as you say, then that makes more sense.

However -- take a look at my test results: it seems to depend very strongly on whether I'm running 2 streams or 8 streams. I have a dual WAN modem (thus 2 up and 2 down should saturate it). The results should not depend on the # of streams I'm using (as long as # of streams >= # of WAN connections) and yet it does.

Seems problematic to me.

sashwa
Mod
join:2001-01-29
Alcatraz
446.4 4.0

sashwa

Mod

xmd, Try selecting all the servers EXCEPT for San Jose.

I'm using the same browser as you and on the 8th the speedtest stopped working for me. Justin had me do that and it's now working. I left everything else blank on that page. And now it's working for me again.

See if that works for you.