I'll say again if the issue was congestion control as you call it on our servers that would affect all service providers....

there has been complaints from non VM users on high speed lines as well and some infinity users if I remember right. I would exclude adsl providers as typical speeds on adsl are low anyway.

congestion control is a funny one, it can affect different connections in different ways so wouldnt necessarily affect everyone.

usually hard proof of congestion is when off peak is fast and peak is slow, when off peak is also slow as well then that to me suggests something else needs looking at. I dont think VM's peering is that bad that it also slows down at 6am, tbb is the only place I see that.

http://en.wikipedia.org/wiki/TCP_congestion_avoidanc...

Edited by Chrysalis (Tue 09-Aug-11 18:05:12)

In reply to a post by MrSaffron:

If it was the code then would show up the same issue on all providers with faster speeds....
A slow PC may have an effect i.e. java not running as well, hence we do a flash version too.

Java is a potential source of problems .. most tester s being flash.

seb

Infinity - was not us again, something odd was going on, and very few people actually posting the traffic pattern so that you can see if its a spiky traffic or TCP ramps being hte issue.

Perhaps email seb and john and explain the tcp congestion theory on their server and see what they say

In reply to a post by seb:

Java is a potential source of problems .. most tester s being flash.
seb

Your Flash tester gives me the same sort of results.

Comparison this morning - http://tinypic.com/m/ete5hz/1

As I said something is wrong and it onlg goes wrong with TBB testing - evrything else runs at very close to headline speeds.

The speedtest.net it is slow for a while at the start, almost as if something is detecting the test and then letting it run fast.

What speeds do you get if you download

1. A single file from http://www.thinkbroadband.com/download.html
2. Two files at the same time from http://www.thinkbroadband.com/download.html

OK I've done that plus a couple of other tests.

http://i54.tinypic.com/2yzncd3.jpg

To test I used jdownloader. I was remote desktop to my home machine. The connection has some other stuff that may or may not have been running - I have applications that upload some data to the web and there is an IP camera plus a webcam running on this machine which can be accessed from a website. Neither gets a heavy load though - maybe a dozen or so visits each a day with an enforced limit of just a few minutes.

The first is 1 x TBB file 1 connection. This is variable but I suppose the average is 1.5 MBps

The second is 2 x TBB files 2 downloads 2 connections (that will actually work out at 4 total connectuions) - say 2.7 MBps

The third is 3 files from Filesonic 12 connections (36 total connections and how I'd normally download) I throttled it back and then cancelled the download at the end. Not sure what the small spikes are at the start maybe there was a hiccup starting or possibly finger trouble - I don't remember. I'd put this at average 5.5 MBps during thegenuinely active period.

The fourth is 2 TBB files 12 connections - say 3.8MBps until the first file completed and then maybe 2.3MBps

Clearly the number of connections have an impact but TBB is performing worse than FileSonic.

Edited by kwikbreaks (Wed 10-Aug-11 15:20:39)

Still could be ISP and peering based, all it proves is that multiple threads are needed to saturate your line and speedtest.net does two threads in its test, we usually do 1.

The ISPA testing used two, and tbbmeter lets you have up to 5.

Reason I say ISP is this: During ISPA testing we did tests that used single thread and dual thread, and on some ISP's no difference, on VM there was a difference that you could see, then with no config changes at our end sometimes you would see a big change in the average.

How different are the routes between tbb and the fillesonic files?

You have probably hit the nail on the head with the number of connections. On reflection given the way cable upstream operates with requested upstream slots a single connection is probably always going to be slower than multiple connections. If Ignition sees this thread he's probably best qualified to comment. Plus even multithreaded the TBB downloads didn't exactly fly.

I think I'll mark that down as probably being the answer rather than congestion at 6:30 a.m. Of course it still doesn't alter the fact that this makes the TBB speedtest useless as a diagnostic aid to me.

> How different are the routes between tbb and the fillesonic files?

Completely different beyond the initial hops....

C:\Windows\System32>tracert speedtest.thinkbroadband.co.uk

Tracing route to speedtest2.thinkbroadband.co.uk [80.249.109.154]
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms 192.168.0.1
2 10 ms 7 ms 7 ms cpc1-nrte22-2-0-gw.8-4.cable.virginmedia.com [82.5.20.1]
3 9 ms 7 ms 7 ms nrth-core-1b-ae1-2555.network.virginmedia.net [213.106.255.209]
4 8 ms 7 ms 7 ms nrth-bb-1b-so-511-0.network.virginmedia.net [212.43.162.241]
5 11 ms 11 ms 17 ms brnt-bb-1a-as2-0.network.virginmedia.net [212.43.162.217]
6 13 ms 10 ms 10 ms brnt-bb-1b-ae0-0.network.virginmedia.net [213.105.174.226]
7 9 ms 10 ms 11 ms brnt-tmr-1-ae5-0.network.virginmedia.net [213.105.159.50]
8 13 ms 11 ms 12 ms telc-ic-1-as0-0.network.virginmedia.net [62.253.185.74]
9 14 ms 10 ms 11 ms linx-gw1.thdo.ncuk.net [195.66.226.240]
10 11 ms 12 ms 11 ms speedtest2.thinkbroadband.com [80.249.109.154]

Trace complete.

C:\Windows\System32>tracert filesonic.com

Tracing route to filesonic.com [78.140.163.15]
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms 192.168.0.1
2 17 ms 7 ms 7 ms cpc1-nrte22-2-0-gw.8-4.cable.virginmedia.com [82.5.20.1]
3 8 ms 9 ms 7 ms nrth-core-1a-ae1-2555.network.virginmedia.net [213.106.255.81]
4 10 ms 7 ms 7 ms nrth-bb-1a-so-210-0.network.virginmedia.net [212.43.162.237]
5 12 ms 10 ms 11 ms manc-bb-1b-as4-0.network.virginmedia.net [213.105.64.22]
6 13 ms 14 ms 11 ms know-core-1b-pc200.network.virginmedia.net [195.182.178.150]
7 12 ms 11 ms 11 ms wb7301a.network.virginmedia.net [62.30.0.204]
8 15 ms 15 ms 16 ms brhm-bb-1a-ge-720-0.network.virginmedia.net [62.30.249.46]
9 21 ms 18 ms 19 ms 0.ae1.mpr1.lhr1.uk.above.net [213.161.65.149]
10 14 ms 14 ms 15 ms ldn-b5-link.telia.net [213.248.104.37]
11 115 ms 36 ms 16 ms ldn-bb1-link.telia.net [80.91.252.205]
12 24 ms 23 ms 23 ms adm-bb1-link.telia.net [80.91.253.17]
13 22 ms 24 ms 22 ms adm-b4-link.telia.net [80.91.253.176]
14 31 ms 31 ms 31 ms ic-130057-adm-b4.c.telia.net [213.248.93.130]
15 29 ms 32 ms 31 ms v-2-fg09-d861-15.webazilla.com [78.140.163.15]

Trace complete.

C:\Windows\System32>

In reply to a post by MrSaffron:

The speedtest.net it is slow for a while at the start, almost as if something is detecting the test and then letting it run fast.

What speeds do you get if you download

1. A single file from http://www.thinkbroadband.com/download.html
2. Two files at the same time from http://www.thinkbroadband.com/download.html

for me #1 just went straight to 3.9meg/sec full speed.

so I didnt need to do test #2.

these files are on a different server to what the speedtest runs on.