Re: Is AT&T Fscking With My Connection?

On Mon, 13 Feb 2012 14:30:37 -0600, Sqwertz <swertz@cluemail.compost>
wrote:

>This just started happening when I told them to downgrade my service
>from 3Mbps to 768Kbps:
>
>http://i42.tinypic.com/mrad1g.jpg

Yep, that's the way it works. They don't really throttle the download
speed by increasing the ACK delays. They just time slice it, which is
what you're seeing. Comcast does the same thing with their 1Mbit/sec
service. You get a burst of traffic at about 8Mbits/sec, followed by
a giant delay with no traffice, followed by another burst. It's not a
problem if you have enough buffering, but a few apps (such as telnet)
go totally nuts with the delays.

Why did you drop all the way from 3Mbits/sec to 768Kbits/sec? What's
wrong with 1.5Mbits/sec?

>Are there any other better logging utilities I use to argue my case
>better with them? I have about 10 of these screenshots showing
>half-loaded and stalled web pages in progress (along with the Network
>Gadget).

Not really. Using the built in traffic monitor in Windoze XP and
above might be more familiar to AT&T. The problem is not convincing
them that something is wrong. The problem is finding someone at AT&T
support that understands what's happening, followed by finding
something that can actually do something about it.

>Speed Stream 4100B, AT&T DSL 768Kbps, hardwired to WRT54G and Windows
>7 x64. Wireless connections to XP are the same.

Good DSL modem. Some WRT54G mutations do strange things, especially
with alternative firmware. This is especially true if you use QoS and
have your reserved bandwidth set too high. Try connecting the Win 7
machine directly to the modem, without the router, and see if anything
changes. It probably won't but it's easy enough to do.

Hint: AT&T is evil. Find a different DSL provider.
--
# Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
# 831-336-2558
# http://802.11junk.com jeffl@cruzio.com
# http://www.LearnByDestroying.com AE6KS

On Tue, 14 Feb 2012 01:59:00 -0600, Sqwertz <swertz@cluemail.compost>
wrote:

>I'm never even getting the full 768Kbps at it's peak.

Windoze 7 is suppose to automagically tweak and tune the network
parameters. However, I've seen it screw up. Try:
<http://www.dslreports.com/tweaks>
and see if anything is mistuned. Also try a different computah.

>The top of the
>scale on the Network Gadget is set at 96KBps. And you can see at that
>peak that it just reached it's running at only 70KBps.

Bits and Bytes don't mix well. You're paying for 768Kbit/sec and
getting 559.4Kbits/sec.
<http://i42.tinypic.com/mrad1g.jpg>
It's like that with all DSL speed specifications. My 1500Kbits/sec
service actually delivers about 1350Kbits/sec. My office
3000Kbits/sec service delivers 2500Kbits/sec. However, your
560Kbits/sec seems a bit low. I would expect about 650Kbits/sec. Try
the local ATT speed test at:
<http://www.att.com/speedtest/>
or <http://www.speedtest.net>

>Which would
>still be fine if it didn't drop out for 5 seconds every 7 seconds.

That's not going to go away. That's the way it's done. Same with
Comcast 1Mbit/sec service. You get a fast burst, long delay, fast
burst, long delay, etc. Incidentally, if you connect to Pandora
streaming music service, you'll see that they do the same thing.
Instead of dribbling streaming music to you at about 32 or
64Kbits/sec, they supply a burst of data at full speed, followed by a
very long interval. They rely on your buffering to make things work.

>If
>I throttle my torrent traffic to around 500KBPS it won't drop out, and
>actually gives me HIGHER throughput than trying to run it at maximum
>speed. Irony, eh?

Yeah.

>> Why did you drop all the way from 3Mbits/sec to 768Kbits/sec? What's
>> wrong with 1.5Mbits/sec?
>
>Pricing/unemployment reasons. Not much demand for SCO support/admins
>anymore. Which I'm sure you know :-)

Yep. I have exactly one non-paying SCO Xenix customer left. No SCO
Unix customers. What's left are a mix of OS/X, Windoze, and a little
Linux. I'm diversifying into sewing machine repair. That should be
hint.

1500Kbits/sec shouldn't be that much more per month as 768Kbits/sec. I
would suggest an upgrade.

>I could get by with 768Kbps just fine if it weren't for these drops.
>Browsers don't like this and some websites will timeout if you're
>running less than 500Kbps, it seems.

They shouldn't drop out or dialup modems wouldn't work. Something
else is happening.

>Which is why I'll probably be calling the Evil Time Warner. But not
>after I call AT&T for ***** and giggles. I need to document it for
>their next class-action lawsuit (I got a whole $1.90 out of last
>year's settlement!).

You're lucky to have gotten the $1.90. Many years ago, I was involved
in a class action suit, where the attorneys got about 2 million off
the top. Restitution "expenses" burned most of what was left. I got
an "offer" for a discount coupon for more of the company's crappy
products if I was willing to fill out a huge marketing questionnaire.
Lesson learned... class action suits only work for the attorneys.

>Just bought this one. Last one lasted 6-7 years (and I'm not so sure
>it's really WAS dead)

Well, there is one potential problem. The 4100B had a firmware update
for PBI/SBC/AT&T PPPoE. Make sure you have 1.0.0.53 firmware. See:
<http://www.dslreports.com/faq/amfaq/2.4_Efficient_4100_modem>

Good luck...

--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

Sqwertz (for it is he) wrote:

>>>http://i42.tinypic.com/mrad1g.jpg

> I'm never even getting the full 768Kbps at it's peak. The top of the
> scale on the Network Gadget is set at 96KBps.

You won't. There's about 15% overhead on DSL between the sync speed and the
IP throughput, although you're seeing a greater difference than that.

--
<http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm@ale.cx)
18:08:19 up 33 days, 21:38, 4 users, load average: 0.05, 0.20, 0.26
"People believe any quote they read on the internet
if it fits their preconceived notions." - Martin Luther King

Jeff Liebermann (for it is he) wrote:

> On Tue, 14 Feb 2012 01:59:00 -0600, Sqwertz <swertz@cluemail.compost>
> wrote:

>> Which wouldstill be fine if it didn't drop out for 5 seconds every 7
>> seconds.
>
> That's not going to go away. That's the way it's done. Same with
> Comcast 1Mbit/sec service. You get a fast burst, long delay, fast
> burst, long delay, etc.

I think there is something amiss here. I can understand only getting 560kbps
when paying for 768kbps service, but surely pausing the traffic for 40% of
the time effectively gives a rate of 330k, if he's peaking at 560kbps?

--
<http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm@ale.cx)
18:11:47 up 33 days, 21:41, 4 users, load average: 1.35, 0.54, 0.37
"People believe any quote they read on the internet
if it fits their preconceived notions." - Martin Luther King

On Tue, 14 Feb 2012 18:19:40 +0000, alexd <troffasky@hotmail.com>
wrote:

>Jeff Liebermann (for it is he) wrote:
>
>> On Tue, 14 Feb 2012 01:59:00 -0600, Sqwertz <swertz@cluemail.compost>
>> wrote:
>
>>> Which wouldstill be fine if it didn't drop out for 5 seconds every 7
>>> seconds.
>>
>> That's not going to go away. That's the way it's done. Same with
>> Comcast 1Mbit/sec service. You get a fast burst, long delay, fast
>> burst, long delay, etc.
>
>I think there is something amiss here. I can understand only getting 560kbps
>when paying for 768kbps service, but surely pausing the traffic for 40% of
>the time effectively gives a rate of 330k, if he's peaking at 560kbps?

That's because the Win 7 widget that he's using is a "network meter"
not a benchmark program. The network meter will show traffic as
reported by the NDIS driver, probably using TCP, which requires ACK's.
The benchmark program will probably use UDP which does not require
ACK's. The lower figure for TCP versus UDP is caused by the dead time
on the uplink between data bursts.

--
# Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
# 831-336-2558
# http://802.11junk.com jeffl@cruzio.com
# http://www.LearnByDestroying.com AE6KS

wikihow.com/Test-for-Bandwidth-Limiting-by-Your-ISP