Draft Minutes

Coordinating Committee for Intercontinental Research Networking
June 28, 1997 – Kuala Lumpur, Malaysia


Shoichiro Asano

Michael Behringer

Suzanne Burgess
DynCorp/FNC - USA

Keith Chang
Industry Canada/GIBN - CANADA

Kilnam Chon

Lilia Cruz
REACCIUN Universidad Central - VENEZUELA
icruz@reaccim.ve and icruz@sagi.ucv.edu.ve

John Dyer

Douglas Gatchell

Jan Gruntorad

Shigeki Goto
Waseda University - JAPAN

Saul Hahn

Ted Hanss
Internet2/University of Michigan - USA

Juergen Harms
SWITCH/University of Geneva - SWITZERLAND

Tarek Kamel

Lawrence Law

Mark Luker

David Macneil

Lucia Melo

Kevin Meynell

Tracie Monk

Kees Neggers

Marie-France Remy
Industry Canada - CANADA

Jose L. Ribeiro

Tin Wee Tan
National University of Singapore - SINGAPORE

Stefano Trumpy

Florencio Utreras

Karel Vietsch

· Appoint leaders for the MBONE, Measurement, and Security Working groups. (Chon, Neggers, Strawn)
· Ensure that caching engineers on CCIRN member networks are aware of the Caching workshops and the planned upcoming workshop.  (CCIRN)
· Work toward standardizing CCIRN member network cache logs formats and make logs available for comparative analysis. (CCIRN member networks)
· Coordinate meetings of the CCIRN MBONE and Measurements working groups after the December IETF. (S. Burgess, Working group leaders)
· Distribute information to CCIRN member networks on the I2 October 7 meeting in Washington, D.C. (T. Hanss)
· Distribute information on splitting research and commodity Internet traffic on the CA*net II (D. Macneil, M. Behringer)
· Circulate notes/discussion from the 1989 CCIRN meeting on equitable cost sharing (K. Vietsch, K. Neggers)
· The next full meeting of the CCIRN will be held in Geneva Switzerland in July 1998, with the Asian Pacific CCIRN acting as host. (K. Chon)
A.  Opening
CCIRN co-chairs (Kees Neggers, Kilnam Chon, and Mark Luker) welcomed and introduced participants.  The co-chairs thanked Karel Vietsch for arranging the meeting logistics and Tracie Monk for the writing up the notes from the last CCIRN meeting.  An action from the last CCIRN meeting was the creation of four working groups which have now met twice.  Notes from the working group meetings in December are located on the CCIRN web site (http://www.fnc.gov/CCIRN.html).  The CCIRN working group meetings in Kuala Lumpur were organized by John Dyer, but the CCIRN needs to select working group leaders to oversee the groups.

B.  Caching
Kevin Meynell reported on the proceedings of the Caching working group meeting that occurred earlier in the week.  (Notes from that meeting can be found at http://www.fnc.gov/CCIRN_97_06_caching.html.)  In particular, the working group discussed the TF Cache work done by TERENA, the current work in Asia caches, and discussed the possible obsolescence of the CCIRN Caching working group and whether it replicates current efforts. Before beginning the discussion of the role of this working group, the Latin America reps described its nascent caching coordination efforts.   Since there is no backbone and connectivity between LA countries, caching coordination is done for internal country caching.  For example, Chile has a hierarchy caching setup.

Returning to discussion at the working group on the purpose of this working group, the members decided that the CCIRN was replicating work already underway.  For the past year, caching workshops have been held held yearly (Poland in 1996 and Colorado, USA in 1997) and another workshop of caching engineers is scheduled for early 1998.  Therefore, the Caching working group decided to disband this working group in favor of increased presence and coordination with the Caching workshops and the CCIRN.

Since a workshop is planned for the beginning of 1998, CCIRN members were asked to inform caching engineers in the member networks of the upcoming workshop.  They were also asked to support the workshop by helping, if asked, to coordinate or publicize the workshop.

Tracie Monk briefed the CCIRN on the last NLANR Caching Workshop (June 9-10, 1997).  It was an invitational workshop held in Boulder, Colorado for 45 caching engineers from 13 countries, and 2/3 of the participants were non-U.S.  The key topics explored at the workshop were:
· status of cache deployment globally
· topology and configuration alternatives
· available cache software, specifically the publicly available, SQUID caching software
· economics/pricing considerations - very important for motivating more cache deployment
· research requirement, specifically the NLANR work
· policy and related issues

Digital announced that it was making select cache logs available (see ftp://ftp.digital.com/pub/DEC/traces/proxy/webtraces.html).  Another announcement is that Vern Paxson will be making trace information available (see Internet Traffic Archive at http://ita.ee.lbl.gov/index.html).  Two areas that were highlighted and need more research are quantifying the costs/benefits of cache usage and the copyright issues associated with caching.  A summary of the workshop is available at http://ircache.nlanr.net/Cache/Workshop97/summary.txt.
The next Caching workshop will be within the next nine months and it was proposed to be held in Europe (UK or possibly Warsaw).  Till the next caching workshop, discussion of caching issues will continue on the IRCACHE list on the NLANR site (http://www.nlanr.net).

CCIRN members were asked to work toward standardizing their cache logs formats and make logs available for comparative analysis.

Kevin Meynell reported on the CCIRN MBONE working group activities, and notes from that working group are located at http://www.fnc.gov/CCIRN_97_04_mbone.html.  The Working group concluded that there was no central coordination of MBONE activity and connections were done on a bilateral, ad hoc basis.  Strong coordination and more user friendly MBONE tools are needed to make this technology more widely used. It was decided that the CCIRN would not be duplicating work being done in this area and should focus on MBONE tools and operation.

There are few MBONE tools to detect private tunnels/connections and little operational coordination in this area.  Tools sets such as MERCI (http://www-mice.cs.ucl.ac.uk/merci/) are not universally accepted.  In North America, several different multicast efforts are underway.  The vBNS (http://www.vbns.net) includes multicast training as connections are added to the network and mbone funding is through NLANR.  Multicast technology is also part of the NGI initiative (http://www.ngi.gov).  Canada has similar multicast efforts through its CANARIE (http://www.canarie.ca).  In the European network, TEN-34, there is discussion of moving the structure gradually to optimize multicast technology. Multicast initiatives in the private sector are described at the IP Multicast Initiative web site (http://www.ipmulticast.com/).

It was decided that increased coordination and information dissemination were the best methods to address these problems.  Towards this end, a workshop on MBONE detection tools and increased MBONE coordination was proposed and the CCIRN approved.   The group will explore different funding sources and work with MBONE experts for this workshop (after the Washington DC IETF in December 1997).

E.  Traffic Measurement
John Dyer provided a summary of the CCIRN Working group discussions and the notes can be found at http://www.fnc.gov/CCIRN_97_06_measure.html.  The working group participants discussed the ISMA meeting in May and the work of the TERENA Task Force in measurement (TF-ETM).  This task force is planning to develop toolkits and provide guidance on common measurement approaches.

Tracie Monk briefed the group on CAIDA (http://www.caida.org), which can serve as an intermediary with commercial operators to facilitate network data acquisition and analysis.  This organization’s purpose is different from NLANR which is concerned with Federal Networks.  Current CAIDA initiatives are collaborative development and deployment of measurement tools - OC3MON for flow monitoring and pathchar for hop-by-hop performance analysis.  More information is at http://www.nlanr.net/Caidants/meastools.html.  Other activities include developing visualizations of Internet flows, IPv6 testbeds, and multicast experiments by hosting collaborative opportunities.

The Internet Statistics and Metrics Analysis (ISMA97) was an invitational workshop held on May 1 and 2.  Notes from the workshop are available at http://www.nlanr.net/ISMA/isma97.html.  Sixty-five participants discussed the different objectives of ISPs, end users and vendors with regard to measurements and their collection/analysis.  ISPs measure to plan capacity, assist with network operations, offer a value-added service, and institute usage based billing for network use.  On the other hand end users use stats/metrics to monitor service performance, plan for connection upgrades, negotiate ISP service contracts, set user expectations, optimize content delivery, and usage policy.  Vendors measure to understand the changing configuration of the Internet and to plan for trends in Internet usage.

Given the different uses of measurements, the measurement methods vary significantly.
The Common Solutions Group (CSG - http://www.advanced.org/csg-ippm/) is performing active tests of one-way delay and packet loss along paths between      measurement machines at CSG sites, and  is planning to deploy measurement machines at all 23 of the CSG sites. They would like to place boxes abroad, and would be interested in talking with TERENA.  Other considerations for the CSG are how to integrate the new connection Internet2 sites, and what specific measurements are necessary, how that data will be stored and analyzed.  Another effort, National Internet Measurement Infrastructure (NIMI - http://www.psc.edu/networking/nimi/welcome.html), is examining the salability of measurements, and Lawrence Berkley Lab (http://www-nrg.ee.lbl.gov/) and Pittsburgh Supercomputing Center (http://www.psc.edu/networking/welcome.html) are working on scaleable alternatives, which piggyback on other protocols.  The IP Measurement Analysis (http://www.merit.net/ipma)  effort is a follow on to the Merit Routing Arbiter project and is placing boxes throughout the U.S. to study routing and route flapping.

E.  Security
John Dyer briefly reviewed the discussion of the CCIRN security working group and the notes from that working group are located at http://www.fnc.gov/CCIRN_97_06_security.html. John Dyer also presented a briefing on Security Incident Response Coordination for Europe (SIRCE) project.  Awarded the first of May, 6-8 Incident Response Teams (IRTs) are participating and more are expected, and will be submitting related security information.  The premise behind this project is that European Computer Emergency Response Teams (CERTs) would benefit from shares incident information contribution. It is envisioned that SIRCE will act as a focal point for European IRTs information and will promote the need for IRTs.  Specifically, SIRCE staff (currently four FTEs) will assist new IRTs (i.e., building a web of trust),  establish a secure communication infrastructure, provide information services, and organize meetings.

After successful completion of the  pilot phase, SIRCE will move into a production phase.  During the pilot phase, SIRCE will provide basic incident coordination.  In the production phase, SIRCE activities will expand to include providing 24/7 service, coordination of existing services outside Europe, and educating new IRTs and assisting in establishing peer relationships.  In addition, SIRCE will be able provide personnel to serve on emergency backup teams, and serve as a trusted party for PGP key signing.

Suzanne Burgess gave an update on the FNC Collaborations in Internet Security (CIS) project. Information on the nine technology testbeds (Advanced Authentication, Kerberos, PKI, Fortezza, Digital Signatures, Secure Messaging, Privacy, Secure Web, National Voluntary Testing Labs) and copies of progress reports and workshop notes can be found at http://www.fnc.gov/cis_page.html.  The CIS participants will be meeting in August to begin work on a final project report and discuss follow-on issues/options.  Overall CIS accomplishments to date are:
- Much greater awareness of interagency needs, resources, and progress. "We are all in the same boat and facing mostly the same problems."
- Increased awareness by users of security vulnerabilities/needs within the participating agencies
- Increased awareness of available security resources and techniques.
- Realization that there are feasible alternatives to passwords.
- Increased awareness of good, useful, innovative solutions being made available by several (often smaller) industry partners.
- Almost all of the CIS work has concentrated on authentication as opposed to authorization. There is a growing awareness that security policies need to cover more than just access to resources (files, computers, programs,...). Example: You need to provide proof of successful training to be able to use an online facility.
- Awareness that rapidly changing technology is both good and bad. It is easy to pick the wrong solution or one that will be made obsolete. CIS is a means of providing some collective wisdom to be used for guidance.
- Identification of those things that keep security solutions from being deployed: cost, cross-platform support, and resistance to non-transparentsolutions by the user.
- Greater synergism of solutions being pushed at different agencies. For example, ESnet and ARL using each others' DCE/Kerberos solutions. NIST RBAC will be used at ORNL.
- Greater leverage to sell security to upper management (another road block). It helps to say "the other place is doing it."

A copy of the MS Powerpoint presentation on the CIS project is available at http://www.fnc.gov/cis_presentation.ppt

The CCIRN working group felt that this was another area that the CCIRN could provide some leadership. Discussion at the working group meeting and at the larger CCIRN meeting found a consensus among the group for strong cryptography. It was decided that the CCIRN, working with CERTs and the SIRCE project, could assist in mapping the status of encryption in different countries. To further define this project, the CCIRN working group will schedule a meeting around the ISOC Security meeting (http://www.isoc.org/conferences/ndss98/).

F.  Next Generation Internet and Internet2
Ted Hanss, Internet2 (I2) Applications Director, briefed the group on the Internet2 project.  (His slides can be viewed at www-personal.umich.edu/~ted/talks/970620I2over.ppt)
Briefly, I2 is a university-led initiative to facilitate and coordinate the development, deployment, operation and technology transfer of advanced network-based applications and network services to further U.S. leadership in research and higher education and accelerate the availability of new services and applications on the Internet.

As stated in their mission statement, advanced applications (e.g., high-end tele-immersion) are one focus of this project. I2 applications will use a network that features, on an end-to-end basis, appropriate class of service attributes.  For example, applications must be able to adjust functionality as the network changes.

I2 engineering personnel are working to develop a network that follows the principles enumerated by I2 (e.g., buy rather than build, open rather then close, …) and  support the needs/applications.  The "gigaPoP" is the keystone of the architecture and a description of it can be found at http://www.internet2.edu/html/gigapops.html.

Ted Hanss reviewed the I2 aspirations for 1997 and 1998.  By the end of this year, I2 hopes to achieve uncongested IPv4 best service and participating universities connected to the vBNS. Furthermore, I2 will identify possible I2 applications, network demands, and begin QoS sychronizing.  The following year, 1998, I2 will introduce support for QoS, IPv6, and multicast.

I2 is planning several workshops.  On July 23-24, there will be an I2 Applications workshop (http://www.internet2.edu/html/applications_workshop.html) and on October 8-9, I2 will have its General meeting in Washington, D.C.  The day before (October 7), there will be a meeting of research and education networks in other countries.  I2 would like to identify faculty and research networks around the world that would like to participate in I2.  I2 is already collaborating with the Canadian network, CA*net II and Singapore is exploring possible collaboration with I2.  In Europe, CCIRN members said that I2 is seen as a U.S. project, but are examining a variety of options in relation to I2.

Mark Luker gave a presentation on the Next Generation Internet (NGI) initiative (http://www.ngi.gov) announced by the U.S. President in the fall of 1996.  (The NGI presentation is similar to the one given by George Strawn and is located at http://www.ngi.gov/testimony.html.) Based on the desire to "allow researchers to live in the future, developing and experiencing technologies of very advanced networking", six federal agencies are contributing to this initiative: Defense Advanced Research Project Agency, Department of Energy, National Aeronautics Space Administration, National Institute Standards and Technology, National Institutes of Health, and NSF.

In addition to a Concept Plan, an Implementation Plan on the initiative will be distributed and will contain more details on the three NGI goals.  The first goal is to perform  experimental research for advanced network technologies.  The second goal is to create a NGI fabric and one of the first steps in this goal will be to bring together the FedNets to begin creating the necessary networking infrastructure.   The first NGI network identified is the vBNS.  To accomplish these interconnect, a federal Joint Engineering Team was formed.  Finally, the third goal is  to develop NGI applications that will integrate advanced networking and application technologies. All federal agencies are working on applications.

G.  International Connections
The NSF High Performance International Internet Services solicitation (http://www.cise.nsf.gov/ncri/nsf97-106.html) was recently released, and Mark Luker provided a brief overview of the solicitation and discussed the STAR-TAP facility.  While similar to previous NSF solicitations, this one contains a different AUP from the past NSFNET AUP because in order to qualify, the connecting network must qualify as a high performance network and each institution on the requesting network must be identified.  In Europe, research networks do not differentiate between institutions on the networks, and therefore, might have some difficulty meeting the AUP for the International Connections solicitation.

Another feature of the solicitation is the proposed single connect point at the STAR-TAP (http://www.startap.net) facility in Chicago, IL.  Several Federal Networks and research partners (e.g., vBNS, ESnet, CA*net II) are already connected to the STAR-TAP, as well as several universities and possibly I2. The facility will support high performance connections and has received interest from around the world.  One neutral, connection point facility was chosen to avoid transit over FedNets from one international connection to another.  Law prohibits transit over FedNets to another network, so the STAR-TAP was designed.  The facility is over-engineer so it is uncongested and works on any network architecture and a technical support team is located at the facility.  It is hoped that international high performance networks will apply for a connection at the STAR-TAP.

Mark Luker noted that the total NSF support for this solicitation will not exceed $4.5 million per year.

One of the international networks already connected to the STAR-TAP is CA*net II, (http://www.canarie.ca/c2) which is Canada's next generation Internet initiative. Designed to promote the development of the next generation of Internet networks, applications and services particularly those that take advantage of a Quality of Service (QoS) Internet. The CA*net II initiative closely parallels the U.S. NGI and I2 initiatives.  It is an IP over ATM network and commodity traffic is segregated from research traffic at the university level or regional level before entering the research network.  Institutions connected to CA*net II must have a commodity Internet connection (similar to I2).

Each CCIRN area gave a brief status report on its current international connections.  In the Asia-Pacific region, APAN is following the analogy of the vBNS and Internet2, in that it is a research, not a commodity network.  Current and planned connections include
a T3 Japan to US(due in fall 1997)
a T3 Japan to Korea(due in winter 1998)
a 15 Mbps Singapore to US(due in fall 1997)

In addition to APAN, the following countries have broadband academic networks.
Australia to US(due in summer 1997)
Japan to US(due in fall 1997)

Latin American connections are below:
· In Brazil, the lines owned by the government are filled to capacity. Fiber is being placed to connect cities along the coast of Brazil and this project is scheduled for completion in December 1997.
· Mexico has  a T3 backbone, with two T3 connections to the U.S.  A T3 is planned to Europe.
Other countries in Latin America (Chile, Peru, Argentina, Venezuela, Uruguay, Paraguay, Ecuador, and Columbia) have one or two lines out of the country.

Connections between Latin American countries include:
· Fiber being deployed along the west coast of Latin America from Peru up to the U.S.
· There is a fiber over the Andes connecting Chile and Argentina.
· Chile and Bolivia are connected.
· Currently, two country PTTs are working to establish a primary backbone in the  southern cone countries.
· Within Central American countries, the RedCACyt (http://ekeko.rcp.net.pe/RedHUCyT/RedCACyT.html) connects academic institutions in the region and includes links to the U.S. and Canada.  The web site contains more information.

In Africa, there are international links dedicated for  research/academic use in Egypt, Tunisia, and Morocco.

One common statistic of all these networks is that they experience more incoming traffic than outgoing traffic.
H.  GIBN Current Status and Future Plans
At the 1995 G7 meeting, 11 pilot projects (http://homer.ic.gc.ca/G7) were announced to demonstrate the potential of the information society and stimulate its development.  Keith Change discussed the second project, the Global Interoperability for Broadband Networks (GIBN), which was designed to facilitate the global interconnection and interoperability of high speed networks and to foster the development and deployment of broadband applications, products and services.  Its objectives were to lay the ground work for future cooperation and promote effective information exchange, by selecting projects of exemplary nature. All the pilot projects have a  national coordinator from each participating country and a national coordinator committee.  Through a steering committee, countries outside the G7 could participate.

In the GIBN, it was decided that the first phase of the project would experiment with applications before working to facilitate a long-term production network. To do this, the project would not add any new bureaucracy and would use existing programs to meet its goals.  The applications chosen covered a wide range and were demonstrated across continents.  Some were based on satellite and this phase lasted for two years.

The accomplishments from this first phase include:
· developing a AUP,
· performing successful trials and demonstrations of broadband applications,
· accelerated boardband interconnectivity (e.g., stimulated the development of the STAR-TAP project),
· achieved first international interconnection of broadband research networks (e.g., CANARIE/Teleglobe agreement to use CA*net II to facilitate a trans-Atlantic link), and
· generally raised awareness.

Beginning in 1997, the GIBN project is moving into its next phase to encourage interconnection of broadband research and education networks.  The group is redefining its approach and, in this next phase, the GIBN has identified governments and networking initiatives it would like to include.  The list of possible collaborators goes beyond initiatives in G7 countries only.  Issues under consideration for the GIBN project include:
· network architecture (e.g., is it possible to have a STAR-TAP-like facility in Europe),
· AUPs
· funding for research and education networks
· cost sharing principles (e.g., France has completed an informal report on this subject)

Keith Chang invited the CCIRN to cooperate with GIBN participants since both organizations are working towards the same goal.  The GIBN and CCIRN have different but complementary constituency and approaches (government sponsored vs. Grass roots; policy level vs. technical/operational).

CCIRN members discussed this proposal and a further discussion was scheduled following the full CCIRN meeting.  It was thought that the two groups could assist each other.  In particular, since GIBN is an official body, the CCIRN could leverage that connection to get its concerns heard at a higher level.

A.  GIBN Report on Cost Sharing
One conclusion from an informal GIBN report on cost sharing was that there is a difference in the funding models used by governments to fund research and education networks.  Some R&E networks receive government subsidies, and often there is no distinction between research and high performance networks.  Other findings include that there is an imbalance of traffic across international connections, and it is no longer clear who is profiting more from the connection since there is a fragmentation of links.  Currently, there are 12 different links from Europe to the U.S. Also an imbalance exists in the price of connections.  For instance, in the U.S. a connection costs 1/3 of the price compared with a connection in Europe.  A link within Europe costs the same as a link from Europe to the States.  It is hoped that deregulation within Europe will lower some of these costs. If a third of the money available from the NSF high performance international connections solicitation ($1.5 million) were used for European connections, that would still only pay for 5% of the connection cost.

The CCIRN members discussed equitable cost sharing for international links.  The ideal  solution would be a model for cost sharing that started with an evaluation of  the global bandwidth available to decide optimal connectivity.  However, that is not the case, and when connections are made or upgrade, other country(ies) needs are not taken into account.  A similar discussion took place during the 1989 CCIRN meeting where a principle on equitable cost sharing was approved, however, it was not possible to implement the statement.

In the U.S., the government does not subsidize R&E traffic, which confuses the subject.  Some high performance traffic is funded, but it is a very small amount.  It is hoped that the NGI initiative will lead to more rational connection point.

In Europe, there is government subsidy of the research networks and their international connections.  However, the high cost of connections often mean that these networks cannot afford a separate connection for commodity traffic only.

Connections to other countries experience unbalanced traffic.  As stated in the previous section, connections from Latin America and Africa have more incoming traffic than outgoing.  In some instances, the country was able to get a asymmetrical link to take advantage of this imbalance.  There is one caution to this metric is that different time zones must be accounted for.  Furthermore, discussion of traffic imbalances is a much more technical discussion and should be deferred.
I.  EuroPoPs
Michael Behringer briefed the group on TEN-34.  It is a high speed pan-European interconnect facility between the national research networks, and consists of 16 national (and one regional) European research networks with DANTE as coordinating partner.  More information on TEN-34 is located at http://www.dante.net/ten-34.html.

Future plans for the network may include a router or "EUROPOP" in the U.S. to maximize a European - U.S. connection.  Many of the networks connected to TEN-34 either cannot afford their own connection to the U.S. or do not have enough traffic to fill a pipe.  This connection could include a router on east coast, which would try to consolidate some of further connections to commercial ISPs and FedNets.   It was thought that some of the QoS lessons (especially how to split research and commodity traffic) from the Canadian network, CA*net II would be useful to TEN-34.

J.  Meeting close

A meeting between the CCIRN co-chairs and GIBN representatives was planned following the CCIRN meeting and its notes are attached below.

Latin American announced it will apply for full membership.  As soon as the details are finalized, it will be sent to the CCIRN.

The next CCIRN meeting will follow the INET meeting in Geneva, Switzerland and is scheduled for July 25, 1998.

There was some discussion of creating a new working group on QoS issues.  Interested parties should e-mail the CCIRN mail list and appoint a working group leader.