Generation of traffic in "settled" peering arrangement

Sean M. Doran smd at
Mon Aug 24 19:14:53 UTC 1998

John Curran wrote (to someone not really worth bothering with):

|   Clarify...  right now, many organizations with high-speed connections
|   to the Internet pay based on usage (including traffic sent).  Doesn't
|   anyone on a usage-sensitive leased-line connection pay based on the
|   traffic regardless of the "legitimacy" of the hits received?  Isn't 
|   this why we all hunt down SMURFers?

Moreover, the technology to mitigate the damage caused by
an increase in demands on a traffic-generating service is
straightforward.  Several companies have inventive solutions
for doing admissions control and traffic shaping, and several
even more interesting efforts (notably some work on congestion-
sensitive admissions control, which seeks to impose TCP-like
congestion avoidance on fundamentally non-congestion-avoiding
traffic) are being worked upon.

As Scott Huddle noted a number of days ago, a source of traffic
in theory *ALWAYS* has it in its power to control the volume and pattern
of its transmissions.   As right now, most such sources are in
easy reach of practical ways of controlling the volume and
pattern of its admissions for very low (or no) cost, a wise
source of traffic in a sender-pays settlements-based world
will already have been investigating ways of optimizing their
traffic output.

The one nit in John's followup is that in a sender-pays
settlements-based world we do not have to hunt down SMURFers,
or flooders -- providing they pay the larger bills based
on the larger amount of traffic, the affected targets will
have necessary revenue to adapt to such things, either through
building bigger pipes or building systems to stop such traffic before
one's egress interface (to avoid having to pay the next network
along for all the flood traffic).


More information about the NANOG mailing list