Aug 02, 2007: Difference between revisions

From ReddNet
Jump to navigation Jump to search
No edit summary
 
(One intermediate revision by the same user not shown)
Line 36: Line 36:


===Notes from the call===
===Notes from the call===
*Bobby Brownstatus of deployment work, rough plan for the future.  Discussion of where to site new depots.
*Bobby Brown reported on the status of deployment work with a rough plan for the future.  Discussion of where to site new depots.
**Vandy internal deployment is almost done. 10 test depots, 4 test meta-data servers, 10 depots at ACCRE, 2 meta-data servers
**Vandy internal deployment is almost done. 10 test depots, 4 test meta-data servers, 10 depots at ACCRE, 2 meta-data servers
**REDDnet domain with monitoring software has been established; management node.  
**REDDnet domain with monitoring software has been established; management node. [http://monito.reddnet.org/nagious2] More info on the mail list.
Nagio page
**They're working with the VTNA using their campus deployment.
 
**Other locations for depots:
2 depots are being sent to PR; that would be 6TB; if we need more we can add to that.  
***2 depots are being sent to PR; that would be 6TB; if we need more we can add to that.  
 
***UMich 6TB
In working with the Library of Congress, ACCRE has discovered a bug and he doesn’t know where it sits. Some allocations on resource 0 are disappearing. The standard way IBP has been used is to configure separate disks as separate depots, rather than as one depot with 4 resources. Under the latter configuration, it’s breaking, losing slices. It seems to be some kind of confluence of factors, which may include.
***Cal Tech 6TB
 
***Florida  6TB
Various possible sources of the problem were discussed.  
**Where will other depots go?
 
***Maybe SDSC – Terry and Micah will try to grease the wheels here.
It’s not reproducible yet.  
***ORNL is another option.
 
**Hardware being deployed:
Working with VTNA
***Tyan motherboard and chasis. Nothing proprietary, though we're getting them through Capricorn.
 
*** Each 1U unit is 3TB=4 750GB drives
External sites where we’re going to install sights
***3 or 4 at UMich cost $2045 per unit=$700 per TB
1) UM 6TB
*In working with the Library of Congress, ACCRE has discovered a bug and he doesn’t know where it sits. Some allocations on resource 0 are disappearing. The standard way IBP has been used is to configure separate disks as separate depots, rather than as one depot with 4 resources. Under the latter configuration, it’s breaking, losing slices. It seems to be some kind of confluence of factors. Various possible sources of the problem were discussed. A major problem is that we haven't found a way to reproduce it yet.
2. Cal Tech 6TB
* Alan, Larry, and Hunter reported on the status of the L-Store and StoreCore work at ACCRE/Nevoa
3. Florida  6TB
**Storecore is working and fully integrated with L-Store. L-store is now using StoreCore for resource discovering, retrieving an updated list of available depots and their resources.
 
**Santi got a new version of Nevoa explorer and is working with that.
Tyan motherboard and chasis. Nothing. Using 3TB per U 4 750GB drives
**The upcoming L-Store release (date?) will be .8.  The L-store team is hesitant to call it 1.0 because there's no fault tolerance on the metadata yet and the security features haven’t been activated. The security feature doesn’t scale at this point and there’s no user interface for it. The team feels that those features are essential for a 1.0 release.
3 or 4 at UMich
** There are issues with the GUI that Santi is working with.
$2045 per TB, $700 per TB 3TB per depot
**Aislan: We tested the L-Store client with Nevoa explorer and store core at their site (i.e. in Brazil). They discovered some problems and fixed them.
 
**L-store team has implemented a system for doing nightly builds of the the L-store software, which is obviously a good step toward quality control.
SDSC – Terry and Micah will grease the wheels here.
** Nevoa have a java installer for Nevoa explorer. They plan to make and installer for L-Store.  
ORNL is another option.
**Further work:
 
***Further Nevoa integration
 
***Packaging for install
 
***Testing status
(2) Alan/Larry/Hunter:  status of L-Store work at ACCRE/Nevoa
*Micah discussed some of the work that Brian Burke (Jim Plank's student) is doing. He has even/odd coding working with the NFU under L-Store.  
    a) LoRS/L-Store view
*Terry raised the question of how were going to get LoRS-view like functionality for L-Store. This is essential for various presentations that we need to make this fall, beginning Sept. 17. It was suggested that we get a student to work on this in September. Terry has concerns about whether we're working hard but not working smart, i.e. whether this plan accords this task the right priority or will be soon enough.  
 
*Micah talked about his recent email about our approach/theory about content control on the depots in relation to the Digital Millennium Copyright Act (DMCA)
Santi got a new version of Nevoa explorer;
Storecore is working and fully integrated with L-Store; Storecore is providing the depot list,
 
L-Store is scheduled to do a release;
There are issues with the GUI that Santi is working with.  
Discussed L-Store with Brian Burke and he has even/odd code working
Aislan: We tested the L-Store client with N-explorer and store core at there site. They discovered some problems and fixed them.  
 
They have a java installer for Nevoa explorer. They plan to make and installer for L-Store.  
 
L-Store will be release .8.  There’s no fault tolerance on the Metadata and the security features haven’t been activated. The security feature doesn’t scale at this point and there’s no user interface for it.  
 
What’s the status on LoRS view? The plan is to assign a student intern to it.
 
 
    b) Nevoa integration
    c) packaging for install
    d) testing status
 
They’re doing nightly builds
 
Micah talked about his recent missive on DCMA and content control on the depots.

Latest revision as of 08:27, 16 August 2007

Coordinates

  • Time: 10:00EST/9:00CST/7:00PST
  • Conference #:510-665-5437 Meeting ID:7333

Attending

  • Alan Tackett, Larry Dawson, Santi de Ledesma, Bobby Brown (Vanderbilt ACCRE, in Nashville)
  • Dan Engh, Paul Sheldon (Vanderbilt Physics)
  • Micah Beck, Terry Moore (UT)
  • P.R. Blackwell, Diana Gunter (SFAU)
  • ??? (Nevoa, via Skype from Brazil)

Agenda for today's call

Bobby Brown: status of deployment work

  • rough plan for the future. Discussion of where to site new depots.

Alan/Larry/Hunter: status of L-Store work at ACCRE/Nevoa

  • LoRS/L-Store view
  • Nevoa integration
  • packaging for install
  • testing status

Micah: LoDN Status

Paul: annual report outline and work assignments

Dan: status of CMS work, discussion of possible RHIC task

PR: status/plans

Upcoming Events

  • Library of Congress meeting, Sept. 17
  • Meeting in San Diego, Oct. 8th.
  • Supercomputing 07

All other business

Notes from the call

  • Bobby Brown reported on the status of deployment work with a rough plan for the future. Discussion of where to site new depots.
    • Vandy internal deployment is almost done. 10 test depots, 4 test meta-data servers, 10 depots at ACCRE, 2 meta-data servers
    • REDDnet domain with monitoring software has been established; management node. [1] More info on the mail list.
    • They're working with the VTNA using their campus deployment.
    • Other locations for depots:
      • 2 depots are being sent to PR; that would be 6TB; if we need more we can add to that.
      • UMich 6TB
      • Cal Tech 6TB
      • Florida 6TB
    • Where will other depots go?
      • Maybe SDSC – Terry and Micah will try to grease the wheels here.
      • ORNL is another option.
    • Hardware being deployed:
      • Tyan motherboard and chasis. Nothing proprietary, though we're getting them through Capricorn.
      • Each 1U unit is 3TB=4 750GB drives
      • 3 or 4 at UMich cost $2045 per unit=$700 per TB
  • In working with the Library of Congress, ACCRE has discovered a bug and he doesn’t know where it sits. Some allocations on resource 0 are disappearing. The standard way IBP has been used is to configure separate disks as separate depots, rather than as one depot with 4 resources. Under the latter configuration, it’s breaking, losing slices. It seems to be some kind of confluence of factors. Various possible sources of the problem were discussed. A major problem is that we haven't found a way to reproduce it yet.
  • Alan, Larry, and Hunter reported on the status of the L-Store and StoreCore work at ACCRE/Nevoa
    • Storecore is working and fully integrated with L-Store. L-store is now using StoreCore for resource discovering, retrieving an updated list of available depots and their resources.
    • Santi got a new version of Nevoa explorer and is working with that.
    • The upcoming L-Store release (date?) will be .8. The L-store team is hesitant to call it 1.0 because there's no fault tolerance on the metadata yet and the security features haven’t been activated. The security feature doesn’t scale at this point and there’s no user interface for it. The team feels that those features are essential for a 1.0 release.
    • There are issues with the GUI that Santi is working with.
    • Aislan: We tested the L-Store client with Nevoa explorer and store core at their site (i.e. in Brazil). They discovered some problems and fixed them.
    • L-store team has implemented a system for doing nightly builds of the the L-store software, which is obviously a good step toward quality control.
    • Nevoa have a java installer for Nevoa explorer. They plan to make and installer for L-Store.
    • Further work:
      • Further Nevoa integration
      • Packaging for install
      • Testing status
  • Micah discussed some of the work that Brian Burke (Jim Plank's student) is doing. He has even/odd coding working with the NFU under L-Store.
  • Terry raised the question of how were going to get LoRS-view like functionality for L-Store. This is essential for various presentations that we need to make this fall, beginning Sept. 17. It was suggested that we get a student to work on this in September. Terry has concerns about whether we're working hard but not working smart, i.e. whether this plan accords this task the right priority or will be soon enough.
  • Micah talked about his recent email about our approach/theory about content control on the depots in relation to the Digital Millennium Copyright Act (DMCA)