The future of NAPs & IXPs

William B. Norton wbn at
Mon Apr 19 15:44:42 UTC 1999


Sorry for the delay in posting a response; there were issues with the
nanog-post list.  Who created that thing anyway ;-)

At 03:09 PM 4/15/99 -0400, etignor at wrote:
>I would like to know what net professionals think about the future of NAPs &

For what it's worth...I just finished a paper that highlights the trade
offs between the direct circuit interconnect model and the exchange point
interconnection model for ISPs. The paper discusses the operations and
financial models (taking into account the circuit costs, cost of exchange
participation, cost of dark fiber, etc.) and the implications of these
strategies across the # of interconnection participants and bandwidth
utilization between the participants. 

To cut to the chase, the major points from the paper:

1) For ISP interconnection, direct circuit interconnection is financially
attractive for low #s of connections (O(5)) of relatively low bandwidth
(DS-3/OC-3). This is due to the fact that ISPs typically pay only half of
the cost for the direct circuits between each other while they pay the full
freight for the big pipe into the exchange. ISPs may want to use the
exchange for a POP, but barring use like that, if all you want to do is
exchange traffic with these other 4 folks into the foreseeable future, the
direct circuit interconnect model makes some sense.

2) As the bandwidth and # of interconnections grow, the exchange point
interconnection model proves much more scalable for two reasons:  First, as
bandwidth grows between participants, ISPs are able to aggregate
interconnection traffic over increasingly large pipe back to their cloud,
yielding potentially significant economies of scale. The direct circuit
interconnection does not provide for this aggregation since the pipes are
destined to different plances. 

Secondly, operationally, fewer backhoes in an exchange, fewer local loop
providers to troubleshoot, etc.  The greater the dependence on
interconnection, the more hardened you want that interconnection environment.

3) Several of the exchanges also provide for the centralization of content
allowing additional transit sales revenue that potentially dwarf the cost
savings highlighted above for interconnection. The direct circuit
interconnect model doesn't allow for this additional revenue opportunity.

4) The analysis finds that facilities-based ISPs win big since (being able
to exploit WDM technologies) they can seamlessly grow the pipe(s) into the
exchange while the direct circuit interconnection model can't take
advantage of this level of aggregation. 

IMHO, these are the most compelling reasons the IXP will be around for some
time.  I can make the white paper with the plots and data available to you
if you like.

>Given UUNET's inclination to peer only with national backbone providers,
what is
>the future of settlement free peering arrangements?
>Will non-profit, cooperative organizations that require multilateral
peering to
>participate (LINX) survive or thrive going forward, or will we see more for
>profit NAPs run by Sprint and co. run the non-profits out of business?

There are some scale issues that come into play that non-profits have to
grapple with. For example, the resources required to build and operate a
commercial grade operations environment from scratch (building close to
fiber, HVAC, Generators, UPSes, Power, Security Systems, 24/7 Staffing,
etc.) with sufficient redundancy excludes most non-profits from at least
the high end of the market. Part of the challenge is achieving the critical
mass to take advantage of economies of scale to compete. 

Another issue is, even if a non-profit could, what would be the motivation
to expend the capital? Is it in their charter for a non-profit organization
to do so, and why would that non-profit organization want to compete
against a for-profit doing the same thing?  In the US, IRS regulatory
issues pop up too, and the tax exempt status may be challenged if too much
revenue is generated in this new line of the business.

>Will co-operative IXPs proliferate in regional markets?  Will this give small
>ISPs, acting collectively, more leverage with the larger carriers (Exodus and
>GTE )?
>I know these aren't truly operational question, but peering and transit is an
>issue that will have an impact on the international infrastructure and may
>result in increased regulation in the future.  Are there other lists where
>people might have some thoughts about these issues?

The answers to these two are too long (and contentious) to answer here.
Since you're in Maryland anyway, I am chairing a panel (with Dave Diaz,
Bill Manning, and Dave Rand) in a few weeks at ISPCON in Baltimore that
will discuss several of these topics. We'll be focusing on definitions and
highlighting experiences in the field. This might be of interest to you and
we can talk off-line there if you want.


William B. Norton	<wbn at>     +1 650.298.0400 x2225 
Equinix Director of Business Development

More information about the NANOG mailing list