dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
5621

Bicephale
join:2005-09-24
canada

Bicephale

Member

MoDem & Bell Portal BitRates both concur to be false!

Click for full size
GNet GBB2060-Xi DMT Capture (2010-May-12, 15h00)

GNet GBB2060-Xi Rescue Tweak (2010-May-12)
 

I don't know what to think of this but here's the story anyway...

My DSL signal started to fail a few days ago, the CRC Error Rate
went through the roof and reached the stratosphere (...) leaving
me no choice but to tweak my connection. The cure worked but
i noticed that both the MoDem and Bell's "Lantern" displayed the
same BitRates and yet a SpeedTest session showed that these
simply didn't match the final speed records... Most obviously, it's
clear that my DownStream BitRate wasn't 5056 Kbps at all so i'm
wondering! Is it possible Bell's HardWare happens to be defective
or worst: Bell found a way to masquarade such disruptions!...


vitesse
join:2002-12-17
Saint-Philippe, QC

vitesse

Member

are you using a proxy? Speakeasy.net is an USA service provider.

Bicephale
join:2005-09-24
canada

Bicephale

Member

Nope, no proxy.
Bicephale

Bicephale

Member

Click for full size
GNet BB0060B DMT Capture (2010-May-16, 10h30)

GNet BB0060B - Normal Setup (2010-May-16)
 

There's currently no apparent DSL signal degradation and yet i can
sense some bizare smell!...


Mauricio0
aka CigaR
Premium Member
join:2008-12-06
Montreal, PQ

Mauricio0 to Bicephale

Premium Member

to Bicephale
Have you tried downloading from sites like microsoft.com ? They are the best real-world speed test.

Bicephale
join:2005-09-24
canada

Bicephale

Member

That's probably a good idea but i don't have other
records for comparisons, contrary to the tests i've
linked to in the present thread. Additionally, these
sets of three tests actually result from conducting
between three to six or seven tests per server as
an attempt to eliminate marginal scores/aberrations.

I keep only one sample per server in the end but i'm
confident those records are suitable for comparisons
and the last set clearly falls below the usual range,
unless the actual BitRate and that which i can read
happen to be different, that is...

What type of failure would cause the MoDem and its
peer to agree on the official BitRate while the speed
is cut down in half and there's no apparent trace of
a disruption to blame for it!?



Also, it's not even clear if the problem has a physical
or logical origin, to be honest. I'm puzzled!
Vomio
join:2008-04-01

Vomio

Member

It's a conspiracy.
Speed test sites are now throttled by the DPI boxes.

Just kidding, maybe.

HiVolt
Premium Member
join:2000-12-28
Toronto, ON

HiVolt to Bicephale

Premium Member

to Bicephale
Why does your domain show as speakeasy.net?

What is your isp?

Bicephale
join:2005-09-24
canada

Bicephale

Member

I don't know what i'm looking at, as a matter of fact!!

I'm with VelCom, this information emerges somewhere
among the details of the "Line quality" tests which i've
linked to via the "SpeedTest" link of my initial message.

In any case, my point is that the late tests combined
to their corresponding 'DMT' captures just don't make
sense to me: the BitRate should call for a SpeedTest
result around 4150 kbps, not 2400 or less!


HiVolt
Premium Member
join:2000-12-28
Toronto, ON

HiVolt

Premium Member

said by Bicephale:

I'm with VelCom, this information emerges somewhere
among the details of the "Line quality" tests which i've
linked to via the "SpeedTest" link of my initial message.
I guess Velcom acquired some IP blocks that were previously owned by Speakeasy.net, and they either forgot to get the records updated, or are sitting quiet and their users can access US streaming content online.

I bet you could watch Hulu and the like.
jfmezei
Premium Member
join:2007-01-03
Pointe-Claire, QC

jfmezei

Premium Member

If there is congestion somewhere, you won't get your full line speed for transfers.

So f you synch at X, but only get half of that speed in throughut, you should contact your ISP about it.

Bicephale
join:2005-09-24
canada

Bicephale

Member

Hi to you both,

There was a time when i could listen at on-line music
on the US-only Pandora site but this no longer works:
the later recognized my Canadian location, actually...



It wasn't possible to reach my ISP yesterday because
when i call 1-888-310-5002, the number on their web
page, the menu eventually leads to a dead end after
business hours, which are from 9 to 23 on weekdays...

The truth is i sense something is "shrinking" at VelCom
these days and i'm concerned that it won't be possible
to resolve my present problems with Bell; switching to
cable started to sound appealing lately, i must confess.



Anyway, right now what i need i guess is a way to find
out if the speed drops are resulting from some physical
malfunction or just a form of systemic outsourcing. The
fact remains that i can't trust the BitRate readings from
any source anymore and i wasn't used to that so far...

Bicephale

Bicephale

Member

Click for full size
Quiet Disruptions (2010-May-17, 15h00)

In addition, i'm certainly not used to see disruptions
during "quiet" periods where noise causes no issue!!

Have a look at today's 13h30 event...

dslrocker3
join:2002-05-26
Toronto

1 edit

dslrocker3 to Bicephale

Member

to Bicephale
I just did a speed test myself at speedtest/speakeasy.net and got a result of 5332/661Kbps. These results are typical on my 6016/800Kbps sync rate line and I'm a Velcom customer. I haen't been noticing any slowdowns. It could be something specific to the IP addresses that you are getting but that doesn't explain why I haven't got one of the bad IP addresses, if that's the case.
dslrocker3

2 edits

dslrocker3 to HiVolt

Member

to HiVolt
said by HiVolt:

said by Bicephale:

I'm with VelCom, this information emerges somewhere
among the details of the "Line quality" tests which i've
linked to via the "SpeedTest" link of my initial message.
I guess Velcom acquired some IP blocks that were previously owned by Speakeasy.net, and they either forgot to get the records updated, or are sitting quiet and their users can access US streaming content online.

I bet you could watch Hulu and the like.
Velcom IP addresses have been showing up as speakeasy.net (according to broadband reports) for the longest time.

No, Hulu and other US content websites don't work. Canadian content websites do work. It seems that pretty much everyone knows that that these IP addresses are Canadian except for BBR/BBR test sites.

pnjunction
Teksavvy Extreme
Premium Member
join:2008-01-24
Toronto, ON

3 edits

pnjunction to Bicephale

Premium Member

to Bicephale
I would try some other tests, or the same tests on a different (faster?) computer.

I noticed that these java speedtests SUCK on my crappy old laptop (one of the first dual-cores, turion tl50 1.6ghz).

Behold:

Teksavvy speedtests in Firefox and IE



Look familiar?

Now, same speedtest on my shiny new i7 desktop:


Finally if I go to speedtest.net on the laptop:


So it is not the laptop connection, it is the test. Like I said, you might want to try other tests and some other PCs. You could be seeing the same thing I see on my laptop in those tests, but maybe not...

I don't know what the difference could be other than speed, but I just checked and the laptop CPU isn't doing much (10-20%) during the java speed test. Java problem? I might look into it further, but it'll have to wait until tomorrow night. I can access my home PCs via VNC from work, but not do speed tests at the same time.

The other thing you could try is the trick of telneting into your modem/router and running the speed test directly on there sans interface by initiating a file transfer (I think from the speed test file to 'nowhere'). I don't remember how to do this exactly, but it was accurate IIRC.
dslrocker3
join:2002-05-26
Toronto

dslrocker3

Member

As pnjunction was saying, Java-based tests are unreliable. They have always been all over the place and if those tests are the only source a lower test result, I wouldn't even worry about it.

Bicephale
join:2005-09-24
canada

Bicephale

Member

If i'm not mistaking, i've been using these JAVA-based
tools since September 2005 (according to the method
i've described previously) and nobody complained to me
about their unreliability so far. I'm ready to evaluate an
other method but lets keep in mind that i have nothing
else than JAVA-based tests to base my comparisons on.

All things being equal i see no reason that would justify
why i should reject my results all the sudden and it still
fails to explain the disruptions i get in absence of noise!



Those speed drops are not a dream and they don't result
from a BitRate reduction, if the HardWare isn't broken on
Bell's side then i must begin to search for IP-based tools.

So, guys, my question is this: what IP-based tool would
make you happy?
Bicephale

Bicephale

Member

Hummm...

Two regular posters just reminded me that i can switch
to an alternate ISP using a trial LogIn account so that i
will be able to compare my final speeds... Even the DSL
link would remain unchanged, nothing else but the ISP is
going to be different...

What do you say about that guys?

alpovs
join:2009-08-08

alpovs to Bicephale

Member

to Bicephale
Try this: »netalyzr.icsi.berkeley.edu/
Maybe you'll discover other problems.
Are connecting through wireless by any chance?

pnjunction
Teksavvy Extreme
Premium Member
join:2008-01-24
Toronto, ON

1 edit

pnjunction to Bicephale

Premium Member

to Bicephale
said by Bicephale:

If i'm not mistaking, i've been using these JAVA-based
tools since September 2005 (according to the method
i've described previously) and nobody complained to me
about their unreliability so far. I'm ready to evaluate an
other method but lets keep in mind that i have nothing
else than JAVA-based tests to base my comparisons on.
The speed tests just return numbers, as far as comparisons they are limited in terms of repeatability and accuracy anyways. Any changes less than 3-5% are pretty insignificant. For a 5/800 sync, I'd define anything about 4.2/0.65 or higher as a pass and anything below (depending on how far below) as a concern. That's about all the utility there is in speed tests IMO, trying to get accurate comparisons (it's 50 kbps better when my modem is upside down!) instead of just pass/fail with these tests is kind of folly.

I think the java tests used to work properly on my laptop as well. No idea what changed, but you could go nuts trying to fix it instead of just forgetting about it like I did. Unless your ISP or area has problems I'm pretty sure you'll get a 'pass' on a speed test that is working properly (either that one on a different computer or a different test).

As far as evaluating another method have you tried just wandering over to speedtest.net to quickly do a run there? The Montreal server was sometimes better than the Toronto server in months past (the run I posted from the TO server last night was 'perfect'), but they're both decent most of the time I've tried them.

Bicephale
join:2005-09-24
canada

Bicephale

Member

Well, i think it's time i post a link to my initial post:

»MoDem & Bell Portal BitRates both concur to be false!

...
alpovs
join:2009-08-08

alpovs

Member

said by Bicephale:

Well, i think it's time i post a link to my initial post:
What's the point in asking questions here if you ignore the answers/suggestions? Why haven't you tried any other speed tests, such as speedtest.net? So far everybody suggested that your speedtest is probably wrong.

pnjunction
Teksavvy Extreme
Premium Member
join:2008-01-24
Toronto, ON

pnjunction to Bicephale

Premium Member

to Bicephale
Your test history seems 'bipolar' in that some of the groups of 3 runs are fine, and others are bad (looking closer some are bad on upload, some down, some both). No idea what is different between the different times?

I would try another test when it is being bad to see if it shows the same thing to isolate the test. I think runs of 3 different tests might be more useful than 3 runs of the same test, at least to verify that the test isn't the source of the inconsistency.

Bicephale
join:2005-09-24
canada

Bicephale to alpovs

Member

to alpovs
My speed tests are right and you are wrong, what's the
point of posting messages which tend to move away from
the topic? I don't think you understand the situation.
Bicephale

1 edit

Bicephale to pnjunction

Member

to pnjunction
Hi PNJunction,

The "out-of-range" US results are due to "rescue tweaks"
which are necessary to remain connected - which is what
is going on right now!... It's called "Spectral Shaping": the
US band, DS band or both are sacrificed and i documented
this with time-stamped 'DMT' captures.

Three different speedtest runs made of +/- three to seven
trials each is what you get, i'm not sure what you mean but
i'll re-read your post attentively...

My present BitRates are 2720/288 Kbps, by the way... This
is why i intend to proceed with two different ISPs using the
exact same HardWare/SoftWare test configurations.
TierX
join:2009-01-20
Canada

TierX to Bicephale

Member

to Bicephale
Call Velcom and open a ticket.

If your modem says you're trained at 5056kbit, then you're trained at 5056kbit, there is no secret Bell conspiracy going on. One CV/CRC error can blow away at least one ethernet frame, getting a handful in a short period can severely degrade throughput, if you are only getting a handful of CRC errors in a 15 minute interval, I'd obviously its not a physical layer problem.

Have you tried a different modem, tried the speedtest from a different PC ? Can that GNET modem do an oam-ping to the dslam line card, if so do you get cell loss?

Assuming its nothing on your end (which I think is likely, as you don't seem to be idiot) it could be a bad port, bad line card, or simply congestion on Bell and/or Velcom's network.

Bicephale
join:2005-09-24
canada

Bicephale

Member

Click for full size
VelCom - GNet BB0060B Rescue Tweak (DMT 18h27)
Click for full size
Bell - GNet BB0060B Rescue Tweak (DMT 19h31)
Hi TierX,

Thank you for your input. It's been a busy evening but i'm
finally able to post some results, although it seems evident
to me that the half-speed syndrome was gone by the time
i managed to establish a reliable connection via tweaking...

I guess the numbers were normal under the circumstances,
there was some 30 ms improvement in latency when i tried
the same tests with a Bell account but the rest was within
my expectations, as far as i'm concerned.

Maybe there will be a next time, i'd be surprized if not to be
honest.

Bicephale

Bicephale

Member

Oups! I almost forgot:

»netalyzr.icsi.berkeley.e ··· 29f-ac00

»netalyzr.icsi.berkeley.e ··· 100-8c33

I need a break...

alpovs
join:2009-08-08

alpovs

Member

See my post about Velcom's MTU: »MLPPP and fragmented UDP traffic
High network buffering is due to QoS in your router (I presume).

On another note, have you checked the wiring at your NID? Once, I started getting all kinds of modem errors that I never got before. I found the wires and nuts at the NID were corroded. I cleaned them, and no more errors. This usually happens when seasons change, which is now.