Rogers explains 'shaping' policy

[Lennart Sorensen]

On Thu, Jun 12, 2008 at 05:41:55PM -0400, James Knott wrote:
> I've wondered about this too. Ethernet switches generally use some form 
> of "round robin" to ensure all ports can transmit data. This means that 
> if you have a huge pile of data to send, after you send one packet, 
> everyone else gets a chance to send a packet, before you can send a 
> second packet. Now I realize that contention resolution is handled 
> differently in cable modems, but there is such a mechanism to ensure 
> fair access. Why does this not help to ensure fair use? Also, Rogers 
> charges users who go over their allocated data amount. You'd think that 
> might slow down some users. However, I can understand Rogers' concerns 
> about the amount of bandwidth some of those peer-peer services consume.

Many protocols have to wait for the acknowledge before sending another
chunk of data.  Large file transfer generally do not.  So you end up
with so much file transfer traffic waiting already, that you can long
delays between chunks of data on the interactive stuff.

Of course there are ways to help that too, such as the ToS values in
packets, which are supposed to let you prioritize low volume interactive
traffic over high volume bulk traffic, although how many ISPs actually
follow those values (since you have to trust that the user's
applications mark the packets appropriately) I don't know.  Certainly
some switches do obey those values to determine what should be forwarded
first to a given port as do some routers.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists