more on interesting rhetoric from your high speed internet provider

[Dave Farber <dave () farber net>]

------ Forwarded Message
From: John Pickens <jpickens () sprynet com>
Date: Tue, 26 Nov 2002 00:56:17 -0800
To: dave () farber net
Subject: Re: <[IP]> interesting rhetoric from your high speed internet
provider

I agree that the term "thieves" is "interesting".  The article also defines
the term "bandwidth abuse".  These are loaded terms, and certainly capable
of fueling controversy and polarization between "operator" and
"subscriber".  But the article does fairly describe the limitations of
current deployed MAC protocols and also fairly describes solutions that are
available for getting the bandwidth problem under better and fairer control
for everyone - both operators and subscribers.

Looking deeper, there is a "logic of the commons" phenomenon that is at
play here - and there are often few "good guys" or "bad guys" - but
primarily normal human beings accessing the "commons" (bandwidth) but
increasingly hampered by "growth" (of the subscriber and application base)
and "greed" (rational subscriber insists on lowest response time and
demands to increase use of cool applications).

Let's face it.  First mile broadband access is cool.  It is always-on and
is blazingly-fast.  So most subscribers do what is natural - they use
it.  And they become dependent upon it.  But often they do not know the
impact of what they do.  One becomes dependent upon his internet radio
station - but does not necessarily know its asymmetrical 128Kbps bandwidth
consumption behavior.  Another becomes dependent on the Internet for
instant messenger voice/video chats with parents or significant other but
does not know of its 400Kbps symmetrical behavior.  Another sees that her
cable modem light is blinking furiously hours after a friend leaves - and,
after much labor, discovers that her friend had activated a peer to peer
application, but even so is unaware that 1-2Mbps aggregate upstream traffic
is being generated.

What is so bad about peer to peer?  Nothing - in an environment with
sufficient resources (moving target).  Peer to peer is problematic only in
that its communication behavior is atypical from the applications
originally envisioned for many (first-mile) commons.  Its bandwidth
behavior is reverse-asymmetrical.  Applications a half decade ago were
forward-asymmetrical consumers - web browsing.  More recently symmetrical
applications have appeared (VOIP, instant messenger video).  But peer to
peer is reverse-asymmetrical - in the upstream direction - and the longer a
node stays connected the more peer nodes discover that node and amplify the
upstream traffic.  So peer-to-peer is actually exponentially
reverse-asymmetrical.

Cable and DSL and some fiber-to-the-home protocols were originally designed
with the first usage profile in mind - asymmetric downstream applications
use.  Early cable installations were deployed with a high sharing ratio -
1000 users sharing a 25Mbps downstream channel is typical - 200 users
sharing a 2Mbps upstream channel is also typical.  In the early days when
there are so few subscribers and applications, such bandwidth (the commons)
is perceived as blazingly cool - and cool applications have cool
performance.  But as the number of subscribers rises, and as the
subscribers discover even cooler applications, the commons begins to
deteriorate - and often rapidly (25Mbps for one user becomes 25Kbps for
1000 concurrent users).

So the action of the operator is not necessarily to be perceived as a
police action (in managing the commons) but rather an effort to improve the
quality of the commons for all consumers. (Ok, not all operators are alike
... but they do share the same problem of maintaining the viability of the
commons.)

So the real discussion is about tools - installing more bandwidth (growing
the commons, but at a $$$ cost), installing more proactive MAC layer
mechanisms (like DOCSIS 1.1 and DQOS for Cable) to managing quality of
service more fairly, and installing tools such as the port-swapping flow
detectors and traffic shapers mentioned by the article.

It's an engineer's problem (and solution).

J

------ End of Forwarded Message

-------------------------------------
You are subscribed as interesting-people () lists elistx com
To unsubscribe or update your address, click
  http://v2.listbox.com/member/?listname=ip

Archives at: http://www.interesting-people.org/archives/interesting-people/