Attendees: Slava Ilyin, Alberto Santoro, Harvey Newman, Sylv Ravot, Yukio Karita (by video), David Williams, Shawn McKee (by video), Olivier Martin, Richard Hughes-Jones (by video), Richard Mount (by video), Les Cottrell (by video), Warren Matthews (by video), Heidi Alvarez (by video), Denis Linglin, Denis Gregory, Julio Ibarra (by video), Daniel Davids.
See ICFA Standing Committee on Interregional Connectivity
See Asia
Japan is moving to 10Gbps for SuperSINET to New York by 1Q 2004. In New York,
Supersinet will have a shared 10GE path to GEANT,
Concerning APAN, there isn't any particular change.
See CERN & DataTAG Project Update
Dante started to offer 10G access to the GEANT backbone, the first country to use this has been Italy (GARR), more will follow including Switzerland and therefore CERN (who should get a 10GE access dedicated to R&D by the end of this year). More surprisingly, Hungary and Poland will also be among the first countries to connect to 10Gbps. Some NRENs are moving to dark fiber, others (e.g. DANTE) are rather pushing for traditional wavelengths as they are not convinced by the economics of a pan_European darf fiber based backbone. The economics depend on many factors including the number of hops. The Europeans appear more optimistic with regard to 40Gbps and expect it to be available in 3 years compared to a 5 year U.S viewpoint (expounded by Doug van Houweling of Internet2).GEANT (GEANT2) next Generation may include a testbed.
TransLight is funded for one additional year and the
project has good
chances to be extended. In this context, the NSF funds the new
connectivity between GEANT
and
UKlight is starting: 10 Gbps to
Optical metropolitan testbed are being deployed. Netherlight is testing a Pure optical Calient switch.
New proposals in EU:
- GARDEN: led by Cisco, IPv6 infrastructure, flow
accounting at Gbps
speed, Gbps firewalls, optical network.
- GRANDE: T-system will donate High-speed
optical links
CERN is involved in both proposals. If funded, projects should start
by the second part of
next year
The transatlantic Chicago-Geneva link will be upgraded at 10 Gbps in September. (Merge of the 622 Mbps and 2.5 Gbps circuit)
Recently have achieved close to 1Gbps between Starlight and Japan through CERN, GEANT (London, New-York, Tokyo). The stability of the connection is not very high as these tests have been carried out on a 400ms RTT path implemented mostly across a shared network infrastructure, hence the importance of using TCP/IP stacks such as FAST having improved resilience to packet losses.
Datatag:
SLAC traffic has been moved so it mainly routes via ESnet (rather than Internet2). Now have 300-400Mbps peaks. There was an ESnet/DoE workshop in Reston (near Washington) recently to make the case to OMB for the future of ESnet. Consensus was to have a 3 -tiered network with a production, a testbed, and a network research components.
Building (with Iosif Legrand and Internet2) a client to run on Windows & Linux to allow measurements and archiving access to data. Hope to have a testing application with a database by this summer.
Had 1Gbits/s between Lyon and CERN. For 2004 GEANT has 10G between CERN. Expect RENATER to have higher bandwidth between Paris and Lyon in 2004. Looking at moving from Lyon CERN link to Lyon Paris (RENATER) link since it may be cheaper.
Richard Hughes-Jones was unable to send voice so it was almost impossible for him to
input information. Next
week, Richard will attend to a meeting about networking in
This report needs to be updated to provide an up-to-date viewpoint. New things include 10GE and the difficulty of getting to 40Gbps will lead to commoditizing of 10Gbps and need figure out to use multi 10Gbps bonded circuits rather than 40Gbps for some time to come (big discussions (arguments) as to when 40Gbps will be available), cheaper switch prices, MAN wireless with range of 10 miles.
See PingER & the Digital Divide. Funding for PingER runs out at the end of this year. We are looking for alternate sources. Have contacted NSF but not talked to right people yet, also there are problems with DoE Labs getting funding from NSF. NATO focused on Caucasus and Southern Republics, Hans Frese might be a contact. ICTP hopeful for funding from EU next year. David Williams pointed out that if we were to lose the monitoring it could be embarrassing for ICFA, so they might be a source, and need to be notified. The amount of funding needed ranges from $50K to $250K, the lower estimate to just keep it running, but not add any new reports, not improve or make more robust. On going costs are mainly to contact remote sites and monitoring sites when we lose connectivity for a long time, and work with the (or find new) contacts to restore connectivity or find a new site to monitor. Harvey pointed out it is important to have measurements available from more than one site. It would be very easy to extend the monitoring at CERN to cover the Digital Divide sites monitored from SLAC. We would also like to add a second (or 3rd) DD monitoring site at ICTP, but they want to wait until they have funding before proceeding.
Showed a new book on "Digital Divide" by Benjamin M. Compaine. There is a Committee for the Democratization of Information in Brazil.
Procket is a new manufacturer of high speed routers for the Internet. Uses custom chips to get higher density, ground up redesign. Three rounds of funding. Critical mass in 1999. First product based on PC, first ship in 2000. Emphasis on reliable software and hardware, and need to support OC768.
A major driver for multi-cast is the media distribution industry, for example the need to deliver a movie to thousands of movie houses on the release date. This would need reliable multicast tools such as Digital Fountain. Besides the bandwidth reduction improvements another reason is the reduction in cpus needed to serve the data.
IPv6 is a big issue in Asia where there is a shortage of addresses. There is no change in the routing architecture, basically uses IPv4 routing protocols. Hopefully IPv6 will remove the need for network address translation and restore the end-to-end nature of the Internet.
To be held in Rio de Janeiro in February 2004, close to Carnival. Purpose to talk about Grid in HENP particularly for LHC, with an emphasis on the Digital Divide. Maybe 3 days with 6 talks/day (talk duration 1 hour) or 3*1 hour talks + 6*30 min talks/day, i.e. 9 talks/day. Maybe allow half day off to tour Rio. Deadline for registration January 2004. Will be a web page at UERJ. Typical room prices $60-70/night. Typical costs iclude air fare, hotel, per diem and registration come to about $2800/person. Expect about 100 attendees.
See Digital Divide Working Group Website Project Update. An important vehicle is to have a repository of information to raise perceptions and get interest. The suggestion is to have a web site.
See Seranate Outcomes and Follow-up
Look at strategic networking issues for next 5 years. Major partners are DANTE, TERENA (coordinates research efforts for NRENs and universities), plus some academic organizations. The final workshop had a lot of consensus but some areas of concern: how do new countries joining EU provide or use funding (structural funding, i.e. how to invest in weaker economies, used to be investment in roads, we need to ensure some of it goes into networks.) One figure shown indicated is that the cost / Mbit/s / year is no longer dropping exponentially but has stabilized a lot over the past two years. David is concerned about the Digital Divide in Europe itself, e.g. the new countries joining the EU have 4-6 times less bandwidth than NRENs in EU-15 countries. The next countries to join have 20-30 times less capacity than the EU-15. Bosnia-Herzogovina has 5K times less capacity than each of the four most advanced countries. Albania does not yet have a research network. A big cause of the DD is excessively high pricing caused by lack of competition often due to out of date regulatory regime or lack of political will to legislate changes. There can be a reluctance to invest in fibre in small peripheral countries without strong high-tech industry, and also in remote regions of the most prosperous European countries. Actions required include: need to measure the DD inside Europe (availability, prices, what are NRENs doing, performance (SLAC/ICTP); stimulate open access ducting and fibre supply; adapt regulatory regime to wide access to fibre infrastructure at cost-based pricing.
How long does it take to bring say S.E. Europe or Poland into par with the top European countries. Some people do not feel comfortable with the E2E monitoring, the concerns have to do with the detailed validation of the results. Validating the results in detail is extremely labor intensive (communicating with remote site contacts when lose connectivity, understanding pathologies of each site in considerable etc.), and not covered by today's funding.
Still working on improved connectivity to S. America, progress is being made.
The ICFA presentation is in a month's time (coincides with the Lepton Photon conference in August 11-16, 2003). Harvey will need an update on the network monitoring, in particular what is new since last time (e.g. Africa, updates on major plots), plus the generalizing to multiple monitoring sites, starting with CERN (will work with Olivier) but evolving to add a few other well connected sites, and also the need for funding. Les will put together a short report with help from David Williams.
There will need to be an update to the Digital Divide report. If there is a point you want to raise at the next ICFA meeting then get information to Harvey in next 2 weeks.
Next meeting will be early September, will have an update on the ICFA August meeting. Working Group chairs need to come with an outline of their reports.