So most ISP's actually utilise 2 ways of managing traffic? Edge routers to restrict people who have gone over their bandwidth, and traffic shaping/QoS switches to do cleverer stuff like drop P2P packets etc.
Theres various things they can do depending upon their own set up and equipment.
There probably are many switches that do handle this on company networks, I was just talking about how it works at an ISP level.
TBH Im not exactly sure how traffic shaping works and my knowledge is really only based on asking questions on a few occasions with one of the Networking guys - (who is actually considered one of the most knowledgable in the UK about ellacoya profiles). Some other stuff Ive picked up on the way.. and if anything I say is wrong then its more likely to be down to my interpretation rather than what Ive been told.
One of the most common edge routers that ISPs use seems to be Juniper ERX's (several years ago it was Redbacks), but the Juniper ERXs are more powerful and have the ability to perform QoS.
Theres detailed info about using the ERX's for QoS on the enclosed
http://www.m40.net/techpubs/software/erx/erx41x/erx-product-overview/download/software-overview.pdfIn section 3-43 theres a diagram showing 4 different types of traffic queues.
AIUI, most ISPs that use them in this way, tend to use "Gold, Silver and Bronze" prioritisation on their Networks.
Underneath theres some info about how they can also do Policy Management - ie on protocol or from reading the TOS byte.
As previously mentioned are intelligent switches that perform deep packet inspection to tag traffic and allow a lot more profiling than what can be done with just the ERX's.
Whilst the ERXs can pick up traffic based on protocol, they can be fooled if someone uses different port numbers etc. Ellacoyas aren't easily fooled and the packet inspection they do identifies the type of traffic.. and it tags it with the relevant TOS.
-------
A bit of useless info.
From around 2003 BTw started using Juniper ERX's to manage some of their VPs. They also used Cisco, and all was well until the introduction of 1Mb home accounts. Those of us that were around at the time may remember how disasterous it was for many people and how contention kicked in BIG time for the 1Mb'ers.
Basically bandwidth wasnt being shared equally, and whilst a 512kb user may not notice contention kicking in at 400kbps - a 1Mb user would.. and they did!
BT rolled out a nationwide fix in early 2004 at exchanges which took several months to complete. One of the fixes was replacing the Cisco kit with ERXs, to ensure that 1Mb users got their fair share of available bandwidth.
Although BTw dont actually (or didnt) perform true QoS at the exchanges, for a long time I have suspected that they do perform something at a very basic level, because on numerous times when Ive suffered from contention, although Ive seen some pretty poor speeds, latency has never been affected.
In summer 2003 our exchange suffered real bad with exchange congestion and speeds went below 100kbps at one point yet the 30+ users on our exchange reporting problems, not one of them had any difficulty with latency.
