ICFA/SCIC meeting, SLAC, Jun 1, '02

Introduction - Harvey

See http://www.slac.stanford.edu/~cottrell/icfa/ICFASCIC_HBN053102.ppt

Harvey welcomed all. This meeting needs to have lightweight presentations and needs to prepare for what we wish to present to ICFA. We have made the case to ICFA about the importance of networking and the poor connectivity to various regions of the world. We have 3 presentations to ICFA this year and we need to give them useful data and not repeat ourselves. Recent activities include the "Potential of Grids and IT in Romania" in April. 

There has been progress in high speed throughputs from 0.1 - 1.0Gbps, in particular SLAC-Manchester (0.4+Gbps), Land speed record SURFnet to Alaska (0.4+Gbps), Tsunami (Indiana) 0.8Gbps UDP). Tokyo - KEK (0.5-0.9Gbps). Also progress in pre-production & production networking 10MB/s FNAL-CERN (Michael Ernest), 15MB/s disk-to-disk Caltech-CERN (Sylvain Ravot). Progress on the wavelength triangle: CERF-SURFnet agreement, SURFnet secures Lambda to Chicago (L3) & CERN (Global Crossing).

 Teleglobe (UK) will shut down at the end of June '02. KPNQwest files for chapter 11: unknown impact on prices and future planning. CERN has a backup agreement through the Swiss network if KPNQwest disconnects, but the throughput is not as good. Negotiations on the 2.5Gbps lambda DataTag link have been suspended. May take 2 months to resolve. Contract with EU says only have to have it on by Jan 2003. The uncertainty makes planning more of a challenge. May have to fall back to more conventional, and probably expensive providers. Also need to be more conservative and cautious for a while.

Harvey has changed the date for which the US-CERN link needs to move to 10Gbps to 2005, i.e. advanced a bit. The lambda triangle (CERN, Amsterdam, UK) is going ahead according to Cees.  Even though bandwidth goes down, hardware to connect up is going up in cost.

Updates from  members

KEK/Japan - Yukio

http://www.slac.stanford.edu/~cottrell/icfa/yukio-jun02.ppt

DESY - Michael Ernst

http://www.slac.stanford.edu/~cottrell/icfa/michael-desy-jun02.ppt

Russia RBnet has no direct peering with GEANT. The demise of Teleglobe is causing concerns. The Silk road first 6 countries will connect up via satellite to DESY in 2nd half 2002. Cost is about $25K/Mbit/year. 

UK - Richard Hughes-Jones

http://www.slac.stanford.edu/~cottrell/icfa/Rich_icfa_report_Jun02.ppt

There is a proposal to have a demo at iGrid2002 to simulate copying data from a radio telescope to a remote site.

Caltech - Harvey

Caltech has two proposals to connect up via QWest or Level 3. Only after TeraGrid has been going for sometime would they connect up to other nets. Once dark fiber installed then there will be many fibers available.

CAnet4 - Dean Karlen

Moving forward on schedule, starts at end July (when CAnet 3 stops).  Will be a high throughput tests from TRIUMF to Amsterdam to CERN at iGrid 2002 in September, want to move a TByte.

France - Denis

http://www.slac.stanford.edu/~cottrell/icfa/denis-france-jun02.ppt

Renater 3 will be deployed in September. Lyon will have 1Gbps to Renater. Renater has commodity 2.5Gbps to US not for research/HEP. For PP will test a second link to SLAC through RENATER (Lyon - Paris- Franfurt) - GEANT (Frankfurt - DANTE - NY) - Abilene - SLAC. Hope to get 50% of 622Mbps. Will replace RENATER to STARtap link. One year contract so this December. Tried to get dark fiber pair between Lyon and CERN, contacting highway companies, and Geneva Industrial services, problem of last mile in Lyon at the moment. Cost is about 800KEu loan for 15 years plus a small amount / year for 100miles. Current cost is 300KEu / year for a link. 

S America - Sergio

Sao Paola will be trying a demonstration on data transfer for iGrid2002. Someone will be working on it from Rio.

Monitoring - Les Cottrell

See http://www.slac.stanford.edu/~cottrell/icfa/cottrell-mon-jun02.ppt

An important point to note is that with today's network performance we are starting to exceed the performance of disks.

Advanced technologies - Richard Hughes - Jones

http://www.slac.stanford.edu/~cottrell/icfa/Rich_adv_tech.ppt

A big topic is storage virtualization, replacements for TCP such asTsunami, ST, iSCSI. New TFP improvements.

Could group into 2 categories: 

  1. Experimental: such as solve IP stack behavior
  2. Research: e.g. 40 Gbps components

Top choices to focus on were 1Gbps, 10Gbps, IPv6, new transport protocols.

We need to get the information out to the applications users. Must get the high performance networking to the applications. Must involve the production providers.

End-to-end connectivity - Richard Mount

Richard was unable to attend. He was requested to activate this area, could involve last mile problems, have to close the loop on how to get the performance all the way to the application and user. This committee has strong connections to the users and results (as opposed to research on individual components). The component developers are listening since they want to see their tools (e.g. Globus) deployed. Want the working group to cover end-to-end performance, trouble shooting, consulting and help.

Requirements - Harvey Newman

Lot of progress in getting agencies to accept large numbers. Do not currently have a high degree of confidence of predictions at the moment due to technology changes and business failures/consolidation. There is a high degree of uncertainty. May need to provide guidance to the ICFA. May be harder to see if the previously predicted drop in Mbits/$ will work out. Need to project forward conceptually on what we expect to happen. Need to keep our feet on the ground, and make sure achievements of existing experiments (such as BaBar) is fed back into future plans.

Digital Divide - Manuel Delfino

He is about to start work on this. The telecomm company failures is making this tricky. For example, the link to S. America was to be provided by Global Crossing but they are in financial difficulties so unclear what will happen. Manuel hopes to have more spare time to work on this in the near future. The Romania Grid workshop was very optimistic in April with quite revolutionary viewpoints on impacts. We expect that this optimism may be less today based on market conditions. A major problem with the digital divide is getting the connectivity to the institutes. So need to think how we can help with the so-called  "capillarity" problem. Maybe need to focus more on how to trigger concrete improvements. This may tie in the technical support group.

Committees

Monitoring needs to get representatives from the HEP community. For example Olivier Martin to represent CERN, someone for RAL/DL, Yukio suggest someone from Japan. Need a representative from S. America (Sergio will try and get someone), Ilya to provide someone for Russia.

End-to-end connectivity and performance needs to add on end-to-end performance for major production at the Lab, how/if to provide trouble shooting support, consulting, help etc. Need people from virtual organizations (VO) such as CMS, BaBar. We need a representative from FNAL, BNL, CERN, DESY (suggestion from Michael of Martin Gasthuber), Japan (get suggestion from Yukio), Sergio agreed to join this group.

Requirements group led by Harvey includes Larry Price , Charlie Young,  possibly Marcel Kunze from tier 1 center at Karlsruhe (Michael will contact), Peter Clarke from within the UK (Richard HJ will contact Peter).

Digital Divide add David Williams, Sergio suggests adding Alberto Santoro.  

We need to add a global accelerator (virtual organization) network and collaboratory committee. There was a recent meeting on this at http://www.lns.cornell.edu/ganwkshp/. Michael will contact Hans Frese who was on the program committee to see who could give a talk at the next ICFA/SCIC meeting.

General points

For Amsterdam ICFA meeting in July Harvey could give update on committee progress, technical progress, telco uncertainties. Need to behave as if we will go ahead with activities to assist with improvements for less favored regions. Such activities where ICFA and the SCIC can help could include:

Each of us needs to think what we need to be in the presentation for Amsterdam.

Administrivia

Need a meeting just before (2 weeks) the ICFA Amsterdam meeting in July. It will be in CERN. Looking at July 11-12th '02. Will include regional updates, sub-committees' progress, a presentation by Harvey on what he expects to present to ICFA (will include updates on re-negotiated contracts, how severe is the telco market collapse). People should come to the meeting with latest news on contract re-negotiations, latest prices. David Williams talk on SE Europe. Manuel talk on Digital Divide committee and an outline of the report to ICFA, since the report to ICFA will focus on the Digital Divide. Updates from sites.


[ Feedback ]