Rough Notes by Les Cottrell
See http://icfa-scic.web.cern.ch/ICFA-SCIC/
Harvey Newman, Richard Mount, Les Cottrell, Dongshul Son, Richard Hughes Jones, Alberto Santoro, Shawn McKee, Shiela Cisko, Heidi Alvarez, Olivier Martin, Dean Karlen, Yukio Karita, Denis Gregory, Sylvain Ravot, Slava Ilyin
See http://pcstats.cern.ch/icfa-scic/20030527-ICFA-HN.ppt
Next report due to ICFA in February 2004.
See http://pcstats.cern.ch/icfa-scic/20030527-DataTAG-OHM.ppt
GEANT announced 10Gbps access at same price as 2.5Gbps.
Interestingly
As a follow up to the successful high speed disk-to-disk
bulk data transfer from TRIUMF to CERN last Fall, a new testbed has been
proposed, under the leadership of Wade Hong (Carleton University) and in
cooperation with CANARIE. It will comprise of five sites across the country and
CERN, each with a TB scale file server. The servers will be able to connect
directly without routers (end-to-end lightpath) and
indirectly with edge routers (to test user controlled lightpath
provisioning). Initial tests to be performed at 1Gbps are being setup now, with
a goal of moving to 10 Gbps in the future.
Shawn McKee (Michigan) presented a brief overview of the End-to-end work which is starting from the Internet2 E2E TAG (Technical Advisory Group--Shawn is the new TAG chair). The concept is to enable the end-hosts to participate in the network monitoring and testing. Anecdotally, most network performance problems are related to the host or application rather than the network itself. It is important to acquire the hard data to confirm this and use this information to address the most common or debilitating problems.
To do this, we need to allow the hosts to participate in testing when there is a problem. Given the platform independence and ubiquity of Java we are trying to develop downloadable applets (or Java Web Start
applications) which can extract
relevant host information and manage the needed testing (either to specific
service points on the network or to other similarly enabled end-hosts).
The acquired data then can be uploaded to a "central" database service (possible to be deployed by Internet2) for further analysis. By having details of the platform hardware and software we can begin creating problem analysis applications to advise the user as well as helping to define what a given set of hardware and software should be able to support (setting user expectations).
Java will be used to manage the testing applications (the applications will be host specific, not Java). This insures comparable/understandable results based upon known versions of applications. The initial set of tools for testing are ping, traceroute and Iperf.
See http://pcstats.cern.ch/icfa-scic/20030527-Divide-LC.ppt
Funding for PingER measurements runs out at the end of this
year. Want to extend funding. Putting together a proposal for
a small amount of funding from European agencies. Need to identify
Little change since last meeting, network traffic is similar
100-300 Mbits/s on two links, ESnet and Internet2. Also looking at other projects such as National Lambda Rail (NLR).
There is a workshop next week at
FNAL got and ESnet provided OC12
link in the Fall 2003 which is beginning to fill up.
FNAL are about to commission a fiber from FNAL to STARLight.
Want to get up to speed in R&D networking and interested in UltraLight proposal. Vicky will chair a working group
session at the
UKERNA upgrading access links to 10Gbits. Many sites
connecting to MANs at 1Gbits/s Big
drivers are astronomers and HENP. Richard has concerns about the effectiveness
of the links and getting high bandwidth to the user (e.g. end system problems).
UKLight has been approved will provide 10Gbits/s
circuits to
Moscow institutes have good connectivity, 100 Mbps - 1 Gbps. Moscow region:
GLORIAD project is planned to start next year at 10Gbits/s. This could be strategic solution for Russian science and education in part of international connectivity for international projects like LCH. Russian contribution could be the TransTelecom fiberoptic cable sross Russia to China through Siberia.
GLORIAD project (NCSA participating from US side) is a Global Optical Ring connecting Europe, Russia, and China. NSF interested. Was a meeting of NSF, China folks and Russians recently. Some ministries in Russia have signed on. Need to make a proposal to NSF will be part of the new Cyber Infrastructure proposals this Fall. Will cross Russia through Siberia.
http://pcstats.cern.ch/icfa-scic/20030527-DDivRio-AS.ppt
Digital Divide conference in
See http://pcstats.cern.ch/icfa-scic/20030527-Japan-YK.ppt
See
Little change since last meeting, try to keep network
traffic similar ~100
Mbits/s on the two transatlantic links from CC-IN2P3 in
An update would be appropriate since many technology
announcements and new pricing structures. Other things would be the high throughput
testbeds and TCP stacks, MPLS and optical networks.
Will aim for last quarter of 2003 for draft, i.e.
Need to arrange support for IEPM/PingER. Problem with ESnet/MICs is there is no natural place to fit it in since they are focused on research, and PingER is more of a production related activity. Will button-hole people at DoE meeting next year. A HEP lab cannot itself apply to NSF directly, i.e. need university to lead proposal. FIU could be interested. Need to decide on scope, what to include and what kind of funding is needed. Set up conference call with Marvin Goldberger.
Connection to Global Crossing is too costly. GC will see if they can come up with a more attractive offering. Alberto is awaiting results of discussions. There is also a taxes (40%) problem.
Looking for connectivity from
Want to begin to create a new “Culture of Collaboration” in particular at SLAC, FNAL and CERN think about what improvements are needed. At the moment it is a bit unclear how to proceed. We have some excellent reports, but a next step is to get the Labs and collaborations to put more efforts into reaching out into new areas (not just to immediate collaborators). Both SLAC and FNAL support looking at how to do this.