dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
2707

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

justin to xmd

Mod

to xmd

Re: Speed Test Feedback?

do you have any extensions in the browser or any anti virus running?
does the upload report differently on chrome or ie11 or firefox?
xmd
join:2009-06-25
Carlsbad, CA

xmd

Member

Nope - about the same: Download is spot on, upload is very random:

FireFox: »[Cable Speed test: 93.59/1.99 28 ms]
Chrome: »[Cable Speed test: 96.98/5.35 31 ms]

And running a Windows 7 VM in OS X, using IE 11:

IE 11 »[Cable Speed test: 95.96/5.86 49 ms]

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

2 edits

justin

Mod

there is something up with upload, it isnt the test because I see stable uploads for other users trial on trial.

What happens if you set one upload stream?

if i had to guess i would say an anti virus, ad blocker, firewall or other problem getting in the way. The upload is exactly the same as posting a file to a website with a post button. If it is unstable in the test it will be unstable in actual usage. Unless you upload using flash - haha.

Notice here in the log for the slowest upload result:

31.56s - up mbits=6.9 DDDDDDDD [i=3.5 is=6.9]
32.04s - up mbits=6.9 DDDDDDDD [i=3.5 is=6.9]
32.56s - up mbits=7.4 DDDDDDDD [i=8.7 is=7.4]
33.05s - up mbits=7.4 DDDDDDDD [i=8.7 is=7.4]
33.55s - up mbits=5.6 DDDDDDDD [i=1.2 is=5.6]
34.05s - up mbits=5.6 DDDDDDDD [i=1.2 is=5.6]
34.56s - up mbits=3.9 DDDDDDDD [i=0.1 is=3.9]
35.05s - up mbits=3.9 DDDDDDDD [i=0.1 is=3.9]
35.55s - up mbits=2.8 DDDDDDDD [i=0 is=2.8]
36.04s - up mbits=2.8 DDDDDDDD [i=0 is=2.8]
36.55s - up mbits=2 DDDDDDDD [i=0.2 is=2]
37.04s - up mbits=2 DDDDDDDD [i=0.2 is=2]

At 33 seconds, the browser basically stalls, then for the next 5 seconds almost no data is reported as being transferred. This is not correct. If the problem does not happen with ONE upload stream then I'd say that you've found an issue with the combination of any browser + your pc + your modem + servers + multiple uploads at one time.

If you use an OS level ethernet monitor you should be able to watch activity on the interface and see corresponding choke off. Could be a number of things and worth chasing up.

Also your connection is very poor latency wise.

You're in California but to reach your IP address from the *nearest* server to you (LA) the ping time is very high:

traceroute -I  X.X.X.X
traceroute to X.X.X.X 30 hops max, 60 byte packets
 1  border1.te4-4.nuclearfallout-64.lax014.pnap.net (63.251.209.181)  0.463 ms  0.478 ms  0.537 ms
 2  core1.be1-40g-bbnet1.lax010.pnap.net (216.52.255.8)  2.857 ms  3.020 ms  3.079 ms
 3  be4212.ccr21.lax04.atlas.cogentco.com (38.88.197.161)  4.302 ms  4.301 ms  4.300 ms
 4  be2017.ccr22.lax01.atlas.cogentco.com (154.54.0.238)  3.015 ms  4.517 ms  4.516 ms
 5  be2179.ccr23.lax05.atlas.cogentco.com (154.54.41.82)  3.583 ms  3.917 ms  4.055 ms
 6  twcable.lax05.atlas.cogentco.com (154.54.10.250)  11.327 ms  11.150 ms  11.132 ms
 7  bu-ether34.tustca4200w-bcr00.tbone.rr.com (107.14.19.120)  5.957 ms  7.653 ms  7.639 ms
 8  66.109.6.65 (66.109.6.65)  8.675 ms  8.672 ms  8.669 ms
 9  agg1.sndhcaax02r.socal.rr.com (72.129.1.3)  10.250 ms  10.253 ms  10.692 ms
10  tge9-2.crlscaij02h.socal.rr.com (72.129.1.121)  9.586 ms  10.849 ms  10.836 ms
11  tge17-10.crlscaij09m.socal.rr.com (76.166.18.162)  25.235 ms  16.582 ms  29.862 ms
12  cpe-76-176-48-248.san.res.rr.com (76.176.48.248)  28.439 ms  29.495 ms  35.866 ms
13  rrcs-X.X.X.X.west.biz.rr.com (X.X.X.X)  33.746 ms  44.954 ms  51.301 ms
 

So after 6ms I've reached rr.com in california, then I've got to travel 38ms more, within rr.com, to reach you in ... California?

The same story from san jose.

By itself latency is not going to stuff up the results but it might be something you weren't aware of.
xmd
join:2009-06-25
Carlsbad, CA

xmd

Member

I think we are in agreement that something is not right with this overall process, but we are not agreeing on what's wrong - you seem to think it's my setup or network, and I think it's in the test program itself. I suspect we have good reasons.

A new test: I disabled my 2nd WAN so I'm running a single 55/5 TWC cable modem, and I set the test program to use 1 stream up and 1 stream down.

Results:
»[Cable Speed test: 57.15/4.04 25 ms]

Bandwidth Graph:




Although the download graph is smooth, look at the upload : it's bursty and seems periodic.
xmd

xmd to Timmn

Member

to Timmn
Here's a sequence of tests, alternating between »/speedtest and »www.speedtest.net/

Notice how the downloads look great on both tests, but the uploads look kind of weird on both. However, speediest.net always gives a final upload bandwidth # which is what I expect, whereas DSLreports number is all over the map.

These graphs are taken from my Peplink router.



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

justin

Mod

Maybe speedtest.net is recording the highest bandwidth on upload that it sees (closer to the interface, it can, because .. flash). See the little spiky blue graphs? I don't know how you can derive a flat 5 mbit out of those, unless you simply shrug and take the highest spike.

You say the results here are all over the shop, but they aren't, really. 5 megabit is not a lot of upload bandwidth to work with, especially in 15 seconds. Here are your results with one line going (the ~57 mbit results).




They are pretty consistent, and never OVER-read, but read about 4 megabit. You want to see 5 bang on .. unless you want to draw a line across the peak of the blue spikes and say "there it is! that's my upload sync speed!" I don't know how you can expect that. The spikes show that the upload is far from smooth, whether it be here, or ookla. Let me ask you this: see the flat tops of the downloads? why are there not flat tops for the uploads for speedtest.net? at 5 megabit?

I'd do another graph with one of your two bonded lines going but cut off all the crud and do 5 fast tests in a row on either speed test, with as wide an X axis as possible. Lets see this flat 5 megabit. I'm assuming here those blue graphs are nearly instantaneous speed?
justin

justin to Timmn

Mod

to Timmn
 
 
You will probably find this interesting. And maybe if you run more tests from now (the time of this post), you'll get different graphs to those you've got so far.

The screenshots here are with two different server side configurations. The bumpy one is default linux server, or anything more aggressive than default (for receiving data only).

The smooth one is limiting the extent to which the server side receive window can open.

Both tests report pretty much the SAME speed. I haven't tested this extensively, but I pushed out the change so I can look at bulk results.

However the lower server receive window has a problem: it severely caps the congestion window size, so the maximum achievable speed per stream is then limited to a product of your latency to the server. I'm interested to know if this makes testing gigabit UPLOAD speeds more difficult.
justin

2 recommendations

justin to xmd

Mod

to xmd
Ok I have got to the bottom of it.

Not everyone sees the "choppy" upload

The reason for the chop is .. buffer bloat. The server advertises a big receive window, and why not. It must allow someone on fiber to upload at very high speeds over a link that might be 20 or 30ms, so that is a lot of data in flight. Your PC is connected directly to your router and modem and they have very large buffers in them. The PC floods data in, and the latency rises off the charts. Chaos ensues. Someone else may not see this chaos - because their setup does not have huge buffers.

The current config is likely to be reversed

I'm betting you now see nice smooth uploads and no huge latency increases during the upload. Right? that is because the server is saying don't flood data at me. However this is sub-optimal for those on fast connections AND suboptimal for highlighting the problem of buffer bloat in home equipment. By masking it this way, it perpetuates the problem.

Buffer bloat is a real issue

The buffer bloat experts want to see more light shed on this problem. Your setup is showing pretty classic signs. Mine is as well. I've got a BiPac 7800 DSL modem, and it probably has a buffer that is pointlessly large. I'll be looking for the chance to flash it, or upgrade to something different.
xmd
join:2009-06-25
Carlsbad, CA

xmd

Member

Thanks for looking into these issues Justin.

FYI, I just ran the test (after resetting all settings to "default") and got the first results that match what the other tests give me: 96 down/ 9.6 up »[Cable Speed test: 94.71/9.76 29 ms]

I'm not sure what you did, but the results feel "better" to me - perhaps you just tweaked the algorithms to hide all the noise in the graphs? (I suspect the other speed tests must do this) or did you actually find a good solution for the edge cases like mine?

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

justin

Mod

I am glad it is working for you but no changes were made with how the upload measurements are done!

Your old results were really choppy because they were really choppy and stalling.
Your new result is flat because it is actually flat.

Can't say why it is got better but if you have no idea as well, the problem might come back. With two lines and bonding and so on, it might be something that cleared up with a power cycle? Anyway there is more info in the logs now, in the "flow Stats" table so diagnosis will be easier if it happens again.

rchandra
Stargate Universe fan
Premium Member
join:2000-11-09
14225-2105

rchandra to justin

Premium Member

to justin
So would putting your modem in bridging mode and using a different router (such as one with CeroWRT) be an option? Anything supporting the Linux codel queuing discipline would do.

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

justin

Mod

yes thats the normal fix turn the modem into pass through or bridge and get a ddwrt type router.

cshilton
join:2002-12-16
Milford, CT

cshilton to justin

Member

to justin
Glad to see people addressing bufferbloat here.
xmd
join:2009-06-25
Carlsbad, CA

xmd to Timmn

Member

to Timmn

8080 blocked?

I'm the guy with the Dual WAN router (Peplink). After chatting with them, it turns out that the default outgoing bandwidth rule is "lowest latency". As a result, during a speed test, if multiple streams are opened for upload, they will all go to WAN1, until WAN1's ping speed drops lower than WAN2 at which point WAN2 will take some of the load. As one can imagine, this is a fairly random process, which explains why sometimes I get WAN1 bandwidth only, and sometimes WAN1+WAN2.
I just tried the latest test and get a "you can't access port 8080" message - as far as I know, I've got nothing blocking 8080, but I'm not clear on the details.

Justin, can you explain more about what's going on with this message?

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

justin

Mod

If during the upload the browser flunks out and it does so on a port 8080 stream, it gives that message and terminates the test. Of course it may rarely be that something else is wrong.

so i can see the log, check port 80 only in the options, then run a test if it fails at least I can see the log because the error will be error 6.

Barry
join:2008-11-04
Burlington, ON

Barry to Timmn

Member

to Timmn

Re: Speed Test Feedback?

The only issue I have is the speed test notification being too chatty. How many times must I x you?!

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

justin

Mod

Only once.

Unless you fiddle around with browser cookies.

Come on, its 2015, are cookies (especially ones from the site you are on) scary anymore? really?

Barry
join:2008-11-04
Burlington, ON

Barry

Member

I have dslreports.com on my exceptions for CCleaner and it still pops up quite often. Perhaps it's the clear on exit cache setting on Firefox. And yes, yes I am scared.

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

1 recommendation

justin

Mod

I guess so. How inconvenient to lose any personalisation options and so on.

Cookies have really never been, and are not currently, a security problem. And if you don't like to be tracked by 3rd parties, then just target those for deletion. Anything else more complicated tends to just make your general browsing experience more painful and you think, well such and such a site is badly designed how did that get past quality control, but actually it is because you're removing scripts or deleting cookies, etc. At least nowadays that is the case.

raypsi1
join:2001-04-09
Wayne, MI

raypsi1 to Timmn

Member

to Timmn

no results par for the course

chromium must not like this test, browser sat there for 10 minutes no results, I cancelled

Barry
join:2008-11-04
Burlington, ON

Barry to justin

Member

to justin

Re: Speed Test Feedback?

I solved it. I needed to add www.dslreports.com as well as dslreports.com. I have Firefox setup to block all cookies with exceptions for sites I use and use Chrome for the ones that act up.

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

justin to raypsi1

Mod

to raypsi1

Re: no results par for the course

Chrome is one of the most frequently used browsers on the test, with a very low fail rate.
(you ran it ok here albeit a month ago »[WISP Speed test: 14.66/1.36 197 ms] )

Does yours have any extensions? if so, what has it got.

raypsi1
join:2001-04-09
Wayne, MI

raypsi1

Member

I have all the same extensions except I dumped adblock plus
Firefox works this test
I think my last speed test didn't use this fancy graphics
raypsi1

raypsi1

Member

tried agin alas it pulled up the pie chart of locals but still sat there

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

justin

Mod

It is a browser extension, and they suck, because they don't tell YOU what they've blown up.

What extensions do you have.

raypsi1
join:2001-04-09
Wayne, MI

raypsi1

Member

i rebooted the comp and took care of the problem merci beaucoupe for sharing
wssddc
join:2001-08-21
Bedford, MA

wssddc to Timmn

Member

to Timmn

Re: Speed Test Feedback?

I get very slow upload results with Seamonkey:

Seamonkey: 58/3.6 (2nd try 60/4.8)
Firefox: 58/57
Speedtest.net in Seamonkey: 58/61

Windows 7, 64-bit
FIOS 50/50

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

justin

Mod

At the start the test does a tiny test download to determine how good the line is.

Seamonkey timed out for 5 seconds doing this "trial" download, so the test guessed the line was very poor and switched to 1/1 stream. That is why if you look at the result you see 1/1 for seamonkey test but normal number of streams for the other browsers.

mckenna797
Premium Member
join:2004-08-25
Astoria, NY

mckenna797 to Timmn

Premium Member

to Timmn
when I went to use the speed test , I cliked on cable the start test and got a popup saying not satellite, latency to low tried several times and got the same message.
azmarisela9
join:2015-09-21
Peoria, AZ

1 recommendation

azmarisela9

Member

The same thing happened to me also. I tried multiple browsers but still nothing worked.