IP accounting

John Curran jcurran at bbnplanet.com
Tue Sep 22 05:37:04 UTC 1998


At 08:44 PM 09/21/1998 -0700, Chris Cappuccio wrote:
>For everything that has been said about charging for bandwidth, I don't
>see why a simple, low priced per-gigabyte model doesn't work.  If you want
>to go to the ironic side, try using free software to do it all...
>http://www2.empnet.com/ipacct/

For sake of simplicity, let's consider the case of _one_ global network
which wants to charge its customers based on the some semblance of actual 
cost (i.e. ignore peering for the moment).  Step one: hunt down all the
marketing folks chanting "value-based pricing" and lock them in a empty
colo facility.   Step two: determine what underlying resources are most
expensive (hint: check out your international circuits).  Step three:
set the unit price for each underlying resource so that current total
usage of that resource times unit price would cover the current bill
for the resource.  Add in a small uplift to the unit prices to handle
misc resources and margin.

All of the above is easy...   The hard part is when you somehow have 
to measure each customers traffic and determine their unit resource
usage.  Given the relatively discrepancy in local, national, and 
international circuits per Mbs/mile, you may need to categorize all
of the flows into appropriate usage buckets.  This gets non-trivial
fast and diminishing returns approaches very quickly.  Currently,
most US providers simply lump the international traffic in with 
everything else and effectively sell "blended bits".  Total traffic
is easy to measure via SNMP, but most folks are referring to something      
more when the use the phrase "ip accounting".

/John



More information about the NANOG mailing list