dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
761
Cloneman
join:2002-08-29
Montreal

Cloneman

Member

packet loss with speedtouch modem w/ traffic?

Long message. Testing will take 10 minutes of your time. Best turn away if you don't like to read!

I've been doing some extensive testing with packet loss on speedtouch modems (ST516, specifically) versus other modems.

I'm finding some issues with packet loss under certain traffic mixes, I'm wondering weather it's abnormal. I'd appreciate if I could get some people to test with speedtouches (and a couple without)

0) Disable QoS on the router side for consistency

Test 1
Saturate download with traffic from various linux ISOs and windows service packs RESULT: no upstream packet loss

Test 2
Saturated download + small upload (SFTP upload capped to 20kB/s - Filezilla )
Result: Suddenly tons of UPSTREAM packet loss reported

I find this quite strange behavior, which I've now seen on 3 speedtouch modems (st516x2, st780), and which I've tested with the help of others.

I've also tested 3 non-speedtouch modems, with the collaboration of other owners (SS5360, Speedstream 6520, and some d-link). None of the other modems show any packet loss in the stress scenario described above.

Intuitively, it doesn't make sense that a connection with a saturated download would have near-zero upstream packet loss, and adding a tiny bit of upload sends it into the 2-6% packet loss.

Here are some links if you need to saturate your download

http://mirror.cpsc.ucalgary.ca/mirror/ubuntu.com/releases//lucid/ubuntu-10.04.4-desktop-i386.iso
 
ftp://ubuntu.arcticnetwork.ca/pub/ubuntu/releases//oneiric/ubuntu-11.10-desktop-i386.iso
 
http://download.microsoft.com/download/0/A/F/0AFB5316-3062-494A-AB78-7FB0D4461357/windows6.1-KB976932-X64.exe
 
 

Packet loss tests be done using the follow site (requires java VM)

»myspeed.visualware.com/i ··· voip.php

You can only do 5 tests/day on that site, but you can just choose a different server and do more. Be sure to choose a green "pin" so you can see detailed results. The main thing i'm focused on right now is upstream packet loss, which can be problematic for applications, I'm assuming.

Here are my typical results on Speedtouch modem with that test & traffic described above.

16mbit dsl, speedtouch (firmware 7.4.3.2)
Upstream packet loss 10.3%
Downstream packet loss 0.4%

3mbit dsl, speedtouch (ST780wl)
Upstream packet loss 1.6%
Downstream packet loss 0.8%

3mbit dsl, non-speedtouch (SS5360)
Upstream packet loss 0.0%
Downstream packet loss 0.4%

5mbit dsl, non-speedtouch (dlink)
Upstream packet loss 0.0%
Downstream packet loss  0.6 %

Downstream packet loss is very similar across the board (and seems normal, small number of packet drops, downstream is completely saturated, this is very acceptable)

Upstream packet loss, however, is constantly at 0.0% on non-speedtouch modems.

Anyway.. all feedback is welcome. I did not do any comparisons for jitter.
Cloneman

Cloneman

Member

in case anyone is curious, I finished my testing and determined that some modems prefer to drop packets in certain types of congestion, and others "buffer" and increase jitter. Just seems to be the way it is.

Ot Chuckcar
@teksavvy.com

Ot Chuckcar to Cloneman

Anon

to Cloneman
I've noticed the internet is substandard in all of Canada except the east coast. Packet loss you'd never see anywhere in the rest of the world you see in Canada. Bell and Rogers spend nothing to upgrade resulting in a deplorable internet backbone not to mention Bell's absurd routing down to Chicago. Until the Prime Minister steps in and privatizes the internet and changes all the laws and dissolves the CRTC nothing will ever get fixed.

Guspaz
Guspaz
MVM
join:2001-11-05
Montreal, QC

Guspaz to Cloneman

MVM

to Cloneman
The ST516 drops packets when the line is saturated while others don't. We noticed this a few years ago when actively working on MLPPP stuff. Nothing much can be done about it, but it doesn't really have much impact on anything either. In all cases, stuff sucks when you saturate a pipe.
Cloneman
join:2002-08-29
Montreal

Cloneman

Member

said by Guspaz:

The ST516 drops packets when the line is saturated while others don't. We noticed this a few years ago when actively working on MLPPP stuff. Nothing much can be done about it, but it doesn't really have much impact on anything either. In all cases, stuff sucks when you saturate a pipe.

Apparently, "saturating" a line means using 40% of the upload and 60% of the download at the same time...

sucks....

Guspaz
Guspaz
MVM
join:2001-11-05
Montreal, QC

Guspaz to Cloneman

MVM

to Cloneman
It shouldn't do that, no... We only dealt with the 100% scenario.
Cloneman
join:2002-08-29
Montreal

Cloneman

Member

if you ever get a chance Guspaz, put my theory to the test with a TP-link 8816 or a ST516. (this was done with a tp-link, but I've repeated similar test with an ST516