r1 - 14 Jan 2009 - 17:13:52 - RobertGardnerYou are here: TWiki >  Admins Web > EsnetNetworkingNotesJan14

EsnetNetworkingNotesJan14

Background

  • Attending: many people from Esnet, I2, BNL and the participating Tier 2 institutions (missed most of them, sorry!):
    • M. Ernst, F. Luehring, R. Carlson, M. O'Conner, H. Severini, P. Mcguigan, M. Sosebee, R. Gardner, S. McKee, Christian, Zane Gray, S. Youssef, Scott Bradley, Jay Duncan

From: Ernst, Michael 
Sent: Monday, January 12, 2009 10:46 AM
To: Shawn P McKee; 'Robert Gardner'; 'Mount, Richard P.'; Frederick Luehring; 'Saul Youssef'; 'Jim Shank'; 'Kaushik De'; Horst Severini; Richard Carlson; O' Connor, Michael P (ESNET); ctodorov@internet2.edu
Cc: Ernst, Michael
Subject: Tier-1 <=> Tier-2 Networking
 
Dear US ATLAS Tier-2 PIs, ESnet and Internet2 Representatives,
 decent, predictable and reliable network connectivity between the U.S. ATLAS Tier-1 and the Tier-2 computing facilities is an important factor in the facilities' ability to serve the needs of regional researchers in the U.S. for ATLAS data analysis. As the Tier-2 centers are playing a vital role in the distributed analysis process timely availability of data is crucial, which requires massive data replication at a high rate.
 
Thanks to the excellent performance of general IP services as they are provided to U.S. ATLAS by ESnet and Internet2 our needs were served well so far. However, given we will be entering the next phase that adds the much more demanding needs of “chaotic” physics analysis to the existing production requirements we, in collaboration with ESnet and Internet2, have to take a look at options that allow us to satisfy the experiment's ever growing needs at all times and at the required rates.
 
I would therefore like to invite you and experts from your local organization to a phone conference to review the current situation and to discuss options to secure the proper development of a vital piece of infrastructure running at the best possible while affordable service level.
 
Coordinates
Date & time       Wednesday, January 14th   at   4 PM (EST)
Phone               (309)946-5300
ID                     735188#
 
Please let me know in case the proposed date & time does not fit your schedule.
 
Best regards,
Michael   

Michael's slides

Discussion

  • Shawn: should include in the discussion technologies such as lambda station & terapaths for dynamic circuit creation and provisioning of bandwidth.
  • Starting with static circuits make make sense, not mutually exclusive.
  • OU: Zane Gray. 10G circuit out to One Net, that connects to NLR to Starlight. What about latency and delay, what is acceptable, and what to do if it falls out of the acceptable range? (probably - latencies are not that critical since this is large scale data movement; what might be important is how "clean", packet loss rate.)
  • Mike: Traffic-engineered circuits to manlan and starlight - need to consider congestion at the major pops, and meet the universities half-way. Peer directly with the universities to connect directly.
  • Bringing BNL to the starlight. UC for example has an Esnet peering at starlight - make an additional step, peering with BNL at starlight. Then traffic can go onto the "science network" - which is a dedicated, best-effort network at L3.
  • Christian: can also plug together L2 circuits from I2 and deliver to BNL.
  • BU: present at NOX in Boston. Christian's option might be viable, DCN circuit brought down from MANLAN. BU-BNL circuit that would look private, establish BGP. What would it take to make this happen? Leveraging existing infrastructure - university to regional connectors of I2. I2 would provide a circuit connection BNL, easy since I2 and Esnet are mostly in the same places.
  • Michael - can we setup a pilot with a couple sites.
  • DCN demos w/ terapaths at endpoints performed for xfers between BU, UMICH, BNL. Leverage this, bring into production. Note - no fees for DCN from I2.

Action items

  • BU-BNL Internet2 DCN pilot. Email moc@esnet (Mike O'Conner)
  • UC-BNL, Esnet peering; Esnet's MPLS virtual circuit - science network. UC will use CIC OmniPoP to route to Esnet (owns 2 ports, one is general IP the other the science network) using a VLAN. UC to request this to Esnet (Mike). IU global NOC.
  • AGLT2 peers at starlight via the USLHC net ultralight chasis, so it should be straightforward.
  • IU (Jay Duncan) - there are several possible connecting points. Will need to follow-up.
  • SWT2: peering via LEARN in Houston. Presently traffic goes via I2 via NLR to Esnet in Chicago. Need to arrange a meeting with David Caldwell and I2 folks. Patrick will drive this.
  • Mike will be assisting this process and providing a point of contact for US ATLAS.


-- RobertGardner - 14 Jan 2009

About This Site

Please note that this site is a content mirror of the BNL US ATLAS TWiki. To edit the content of this page, click the Edit this page button at the top of the page and log in with your US ATLAS computing account name and password.


Attachments


pdf Tier-1-Tier-2-WAN-Network-01-14-2009.pdf (337.3K) | RobertGardner, 14 Jan 2009 - 16:03 |
 
Powered by TWiki
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback