USLHCNet working group meeting
→
Europe/Zurich
Harvey Newman
(Caltech)
-
-
Introduction - HN
-
Brief site reports
-
1
US LHCNetSpeaker: Sylvain Ravot (Catech)US LHCNet Purchase orders for the 3 new TA links across the Atlantic have been issued by Caltech. Contracts with GC, Colt and Qwest will be signed soon. Transition to the new Caltech contracts and new services from the present services (under contracts with CERN) by November. Transition to optical-multiplexer based core for Layer 1 fallback in case a link goes down, and granular circuit-oriented services.
-
2
CERN/LHCOPNSpeaker: David Foster (CERN)David Foster reported on the LHC OPN meeting, where the agenda shows several presentations that he reports were well received. See http://agenda.cern.ch/fullAgenda.php?ida=a062977 The OPN model of use consigns T2-T1 flows involving European T1s to use the general purpose GEANT infrastructure, which others noted may not be optimal in the context of the LHC experiments' Computing Models.
-
3
ESnetSpeaker: Bill Johnston (ESNet)Bill's Johnston's slides show a clear evolution of the SDN infrastructure with multiple wavelengths on various legs of the Internet2/NewNet footprint. Details subject to funding and loading experience. OSCARs will be used to define paths across this increasingly rich infrastructure. Change from 10 to 100 Gbps wavelengths estimated to occur in the 2010-2011 timeframe. Support for T2 - T1 connections possible, and compatible with ESnet AUP which is "one end must be a DOE facility".
-
4
Internet2 and NewNetSpeaker: Rick Summerhill (Internet2)
-
5
FNALSpeaker: Don Petravick (FNAL)Phil Demar reported Fermilab will soon have 4, then 6 10G waves to Starlight using their CIENA equipment. Also 2 10G waves for development. Fermilab also will put in place 2 1G circuits to CC-IN2P3 using the NSF/IRNC and HOPI waves across the Atlantic (the latter via London). Bill Johnston noted that GEANT may ask ESnet to pay for 1G circuits terminating at DOE Labs at a high price per unit bandwidth since ESnet does not itself operate TA circuits. Ian Fisk reported on recent experience with SC4. Organized transfers have a 2008 goal of 600 Mbytes/sec peak; 25% of that in SC4. Actual peak transfer rates T1-T2 may be much higher. Foster, Johnston and others mentioned that the experiments need to "challenge" and correct the official documents that quote much lower numbers which do not match the actual plans and developments underway, particularly involving the Tier2s.To allow the much-lower estimates in official LCG documents to remain in place was characterized as "dangerous" (for coherent planning, appropriate levels of financial support, etc.). Johnston and HN noted that these figures do not match the ICFA SCIC documents which used as input for the roadmaps, and which are verified by the evolution of flows observed on ESnet. Don's comments: 1) For CMS FNAL is a large T1. The CMS computing model demands that European T2's have good access to datasets at FNAL. US T2's need good access to european T1's. The US's role in the experiment wil be undercut without both symmetries. (What claim will the US t2's have on european datasets if we can not serve them well?) US LHCNET's announced mission includes servicing the european T2's when cross atlantic data transfers occur. While this mission is not supported by the LHCOPN, the US LHCENT mission is larger than the OPN. However My best understanding is that US LHCNET does _not_ transit traffic from the US to european T2's A sample of traceroutes from FNAL to various European T2's showed the traffic flowing from ESNET then across the atlantic to the european T2's. I hope I am wrong, but -- Looking throught the documentation I have, and asking quick after-meeting questions, I was not able to establish that connections exist that enable US LHNET to transit traffic to european T2's. 2) I am glad to hear the discussion abotut AUP's. I understand that the desiderta is an AUP that is simple to state (but, like a terse math proof, might take quite some thinking to formulate). I understand that the ball is in Craig Tull's court to come up with a succint (and perhaps draft) mission statement, which would seed the crafting of the AUP.
-
6
BNLSpeaker: Scott Bradley (BNL)
-
October 23-25 Face-to Face Meeting at Fermilab
-
US LHCNet AUP
-
Update on PerfSonar deployment for the USLHCnet
-
Next Call
Since the ICFA Digital Divide Workshop in Cracow and Sinaia
(Romania) is the week of October 9, October 17 at our usual
time is proposed.
-