Sept 15, 2006

From ReddNet
Revision as of 12:07, 15 September 2006 by 68.32.107.20 (talk) (→‎Technology: Hardware, Software)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Probable Attendees

  • Blackwell (SFA)
  • Hagewood, Assad (Nevoa)
  • Beck, Moore(UTK)
  • Sheldon, Tackett, Pathak, Brown (VU)
  • Swany (UDel)
  • Sean McKee (U. Mich)

Current Agenda (Start time this week, 10EDT/9CDT)

Application Communities

  • A note about communications: Use the list! (Terry)
  • Fusion
  • News from Electron Microscopy (Micah/Chris)
  • AmericaView (PR)

Deployment and Operations

  • Deployment to ORNL (Bobby)
    • Current status is not known
    • If they need another port open, that could be an issue. If they need another port open for the L-depot
  • Yotta depots at Fermi (Alan/Bobby)
  • Status of AMPATH testing (Surya/Hunter)
    • They have 6 real machines and got 440Mbps for about 12 hours. Machines that they're using is shared and also there may be problems with the network configuration.
    • They're getting new dedicated machines and they're about to start testing the link with those. But iperf is getting poor results on those links.
    • Martin is going to get accounts on these machines so that he can look at it.

Technology: Hardware, Software

  • Sean McKee described his efforts to build a cheap box that would give him a GB per second transfers, disk to disk.
    • They've got a boxes with 10Gb links at UM and CERN that they're working with. But various people are working with them.
    • Micah talked about the possibility of putting other data movers (e.g. bbcp) in side IBP. L-store uses the lower levels of the IBP interface and therefore might be able to do this.
    • From the PoV of the LHC computing model, the depots are buffers. They're looking for a host that could serve as a data pump.
    • Mentioned the possibility of using Clyde to test
    • Fermi can saturate a 10G link, but they have a couple hundred nodes on each end.
    • Question of what we can do at SC with this is still open. We want to try to start testing this combination of technologies soon.
    • UM team will be moving to the Tier2 center soon.
  • Current state of the components of the L-Store software Suite(Alan, et. al)
  • Surya on L
    • L-Store
      • The metadata server is ready to go.
      • They are already testing the single meta-data version for testing with Ultralight. They've been hitting the Capricorn hardware and and they're getting 140-150MB per second with 4 depots.
    • Clyde
      • This will be the test harness, based on a Blackboard system. The new version is much more generic.
    • SRM interface
      • It's been completed and shown to be interoperable.
      • The client is on the L-store wiki page. They LBL people have used it to move data from Vandy to CERN. They're using their SRM tester on top of L-Store.
      • Micah wants to set up a persistent SRM/L-Store demo out at ORNL. It would show movement of data via this system back and forth to CERN. Micah will take point on getting this going.
      • Ian Fisk is very excited about this interface, but there SRM version problems.
  • Current state of the Nevoa components and tools (Hunter)
    • StoreCore
      • Have a new version up an running with all the dead depots from the old depots removed.
      • They're moving ahead with advanced IBP clients (reader/writer) using Java. They want a new client version ready for demo.
    • LoDN_Brazil (http://dvnserver.pop-sp.rnp.br)
      • The storage working group is up for review. They're going to show their version of LoDN. It can use either the L-Bone or StoreCore for resource discovery. The integration is done in the WebStart client. There can be user defined volumes added to the store core. This allows people to share a common IBP infrastructure but can also contribute their own. The model they're presenting is a public resource that RNP controls, but which you can add to, but you can also add resources that only.
      • Micah made a point about the problems that we need to watch for with regard to how the different versions of LoDN are evolving manage policy and what that means when you give an exnode to a directory. Hunter's group is doing things like adding group access policy, etc. Need to choose a simple permission structure that people can standardize on.

Organization and Funding Opportunities

  • Status of NSF Physics at the Information Frontier(PIF)
    • Probably not going forward on the PIF.
    • Maybe we could submit it in the spring.
  • Status of NSF PIRE
    • Vandy has given approval to go ahead on the NSF PIRE
    • 6 page preprop due in Oct.
  • What's happening with the SBIR?
    • EoN systems is working on back-up capabilities under L-Store IBP
    • Due Date:

Events, Education, Outreach

  • Update from the OSG meeting (Could be brief)(Paul, Surya)
  • Conversation with Miron Livny (Terry Micah)
  • SC06?
    • Plans for hardware and demonstrations
    • Working with Ultralight
    • Working with OSG/Fermi?
    • Status of Ultralight plans (Alan)
  • Plans for Internet2 Side meeting (Terry, Micah)

Action Items