cancel
Showing results for 
Search instead for 
Did you mean: 

New Samknows Report

Oldjim
Resting Legend
Posts: 38,460
Thanks: 787
Fixes: 63
Registered: ‎15-06-2007

New Samknows Report

Would Plusnet like to comment on this report in one of the blogs http://www.samknows.com/broadband/pm/PM_Summer_08.pdf specially the throttling by port not by content Figure 25 Page 35
Also El Reg link for highlights http://www.theregister.co.uk/2008/08/07/bt_samknows_bandwidth_throttling/
5 REPLIES 5
MickKi
Grafter
Posts: 543
Registered: ‎30-09-2007

Re: New Samknows Report

All I know is that trying to run ssh sessions with servers in the States or Europe most days during the evening is nigh impossible.  After around 01:00 in the morning throttling seems to be slackened and the latency reduces.  Now I wouldn't necessarily confuse ssh with bittorrent - but it seems whoever has set up the Ellacoyas does.  Sad
Not entirely sure whether BT is to blame, but I would love to find an ISP that practices net neutrality in an affordable way.
Mark
Grafter
Posts: 1,852
Registered: ‎04-04-2007

Re: New Samknows Report

I think we've commented on how intelligent traffic management is implemented on our network on many occasions however I'm more than happy to give a rough overview here. Smiley
I did find Sams report quite interesting and I'm looking forward to the continuation of the trials.
Traffic management is not simply by port.
In the Ellacoya rules we have virtually every application you could think of covered.
Some by port, some by IP address, ie source IP, such as game servers, and some by traffic signature, such as FTP, Bittorrent, IRC, VOIP and some games.
Whilst the Ellacoya picks this up and identifies what the traffic is, it then uses that identification against your account type to ensure the correct priority is applied to your traffic.
It reads the signature as a hex string which could look something like this 123456789876543 (very rough indication of a hex) Tongue
The Ellacoya or traffic management looks at the header of the packet, identifies it, classifies it using  Differentiated services (DiffServe) and profiles it.
It's way more intelligent than simply managing by port. Where it simply by port, then the means of circumventing our traffic prioritisation would be childs play and that would defeat the purpose.
Can I point you to the following articles on our portal.
http://www.plus.net/support/broadband/quality_broadband/index.shtml?supporta=qualitybroadband and in particular this article http://www.plus.net/support/broadband/quality_broadband/roadmap.shtml
We've made no secret about how and why we manage traffic on our network. HTTP and time sensitive traffic such as VOIP will always be prioritised.
Hope that helps a bit. Shout if you need further info or clarification.

Oldjim
Resting Legend
Posts: 38,460
Thanks: 787
Fixes: 63
Registered: ‎15-06-2007

Re: New Samknows Report

Mark,
I have a fair idea as to how the plusnet traffic shaping works which was why I was surprised by the result on page 35 where the same file was downloaded as on the test for port 80 but via a different port. I would have expected that there would not have been that sort of difference given that the traffic signature would not have been P2P and it wouldn't have been from the download sites which you restrict or are the restrictions on download sites applied by port number..
Quote
This  test  downloads  files  from  the  same  servers  as  the  Port  80  tests  discussed previously, but instead downloads them from a range of different port numbers that would normally be associated with peer to peer traffic.
LiamM
Grafter
Posts: 5,636
Registered: ‎12-08-2007

Re: New Samknows Report

That is interesting.  It is possible that the Ellacoya kit isn't picking up that particular traffic correctly, or isn't able to identify it at all.  Unidentified traffic isn't prioritised as HTTP is.
oliverb
Grafter
Posts: 606
Registered: ‎02-08-2007

Re: New Samknows Report

Even a very sophisticated recognition system may have some badly written rules.
I remember an anti-virus package that's given me good service for years once reported a file as a trojan horse purely on the basis of its filename.