Revised on 1997.10.08
								Kilnam Chon

Cost Sharing and Intercontinental Research Network Exchange - APAN Perspective
-----------------------------------------------------------

Several issues are being debated among high performance/broadband research
networks around the world including G7 GII/GIBN Project.  Some of them are as
follows;

(1) Intercontinental research network exchange 

(2) Cost sharing among the research networks

1. Intercontinental Research Network Exchange

1.1 Models

There are three models mentioned at GIBN and other meetings;

(a) Single exchange point[Figure 1]

(b) Distributed exchange in North America[Figure 2]

(c) Exchange in each continent with intercontinental connections[Figure 3]

1.2 Model Description

The first model is based on the single global exchange point such as STAR TAP
in Chicago, which was proposed by NSF.  The second model, the distributed 
exchange in North America such as the exchanges in East Coast, West Coast,
and North Border(and possibly South Border).  The last model is the exchange
in each continent with interconnections of the exchanges.

1.3 Assessment

The single global exchange point has the following advantages;

(i)   It is workable now as implemented by NSF, called STAR TAP in Chicago.
(ii)  There are no "transit problems" since all cells packets are exchanged at
      one location.  In particular, this approach solves Asia-Europe link
      optimally until we have the shortest Asia-Europe link, the trans-Siberia
      link.
(iii) It may be most cost effective.

It also has the following disadvantages, too;

(i)   It may cause additional delay.  For example, the Asia-Chicago-California 
      link may have much more delay compared with the direct Asia-California
      link, and it could be serious for some applications, and we need to study
      on this delay*throughput issue.
(ii)  It is not scalable compared with multiple exchange points.
(iii) It is not robust as single failure at the exchange could cause substantial
      damage to the global research network.

The second model based on the distributed exchange in North America is opposite
to the first model with respect to the advantage and disadvantage.  The third
approach of the exchange in each continent would be the long term issue if it
ever be realized since we have to develop the exchange in each continent and
come up with the funding for the intercontinental links.

1.4 Recommendation

Our recommendation is to start from the first model with STAR TAP in Chicago
as the single global exchange in 1997-1998.  All (high performance) research
networks are expected to link to STAR TAP or have transit arrangement to STAR
TAP.  We also start working on the development of other global exchange
points possibly starting from the one in the east coast of North America.  The
exchange at Perryman could be considered for this purpose.  The transit between
these exchange points need to be coordinated among the concerned parties 
including possible Asia-Europe links.  It may also be coordinated with the
transit in other continent.

Since many broadband research networks in Asia and Europe are linking to North
America in 1997 and 1998, the coordination with an appropriate action plans 
should be developed now.


2. Cost Sharing Among Intercontinental Connection of Research Networks 

2.1 Description

Networks are classified differently between USA, and Asia and Europe.  In USA,
there are

	commodity Internet, and
	research network for high performance applications.

In Asia and Europe, there are

	commercial Internet, and
	academic(or research) network.

2.2 Assessment

All commercial Internet in Asia and Europe are considered to be commodity
Internet in USA.  The academic networks in Asia and Europe are typically 
commodity Internet, and high performance research network in some cases.  
See Figure 4 for diagrammatical description.  Thus, we need to watch out 
when we talk about "cost sharing".  

APAN and some other research networks in Asia and Europe are the exception,
which could be discussed on the "cost sharing" with the research networks in
USA.

We have to watch out the current major development of the high performance
research networks such as vBNS, Internet 2, and CA*net 2.  They may have
major impact on the definition of high performance research applications 
compared with the commodity Internet traffic, and other issues such as the
GigaPoP.

2.3 Recommendation

The academic networks in Asia and Europe needs to discuss on commodity
Internet traffic with commercial Internet service providers in USA, and Asia
and Europe.  This may be coordinated with other activities on the Internet 
settlement such as the one for the commercial sector.

Regarding the high performance applications, we need to coordinate with NSF and
other US agencies.  Other expenditures in addition to the communication cost
may be taken into consideration as commented by our counterpart in USA.  
Examples are the R&D investment such as Next Generation Internet Initiative in
USA, and the operation cost of the exchange point and others.