Workers Section

Milestone: Demonstration at TTF (April 3-6, 2002) and Sherwood (April 22-24, 2002) Fusion Science Meetings:
TRANSP analysis code running on the PPPL Linux cluster working on C-Mod and DIII-D data.

Milestone: Demonstration at SuperComputing 2002 (November 17-22, 2002) and APS/DPP (November 11-15, 2002)
What is our focus here?

Implementation Philosophy Focused on April 2002

• Focus is on TRANSP but our implementation is very general so other codes have a roadmap to be integrated into the Collaboratory environment
• Create TRANSP remote invokation environment - general so it can be used for all tokamaks
• GUI to organize data preparation to be data driven so it is general
  • Data driven means, for example, text nodes that indicates what code to run or pointers that indicate where to get data
• If GUI implementation is general, the model can be used for other codes
  • Other codes can be any other code not just power balance codes
• IDL-based visualization will be Power Balance Code specific as opposed to TRANSP specific

• Invoked from showroom using GUI
• MDSplus secured via Globus; what about SQL?
• Use MDSplus Server at DIII-D, Linux cluster at PPPL
• Demonstrate some data preparation, not all will take too long
• Create sample TRANSP run that is short for a demonstration
• TRANSP runs loads MDSplus data and updates Run Management Database
• Simple visualization (IDL or OpenGL?) on showroom floor driven by MDSplus event

Working Plan for April 2002

1. Linux Cluster at PPPL
   1. Hardware [PPPL, 2 wks - Done]
      1. Hardware delivered
      2. Hardware assembled
      3. Hardware tested and certified
   2. Globus Software [PPPL, 2 wks - Dec 21]
      1. Installed on cluster
   3. Batch Queue Software [PPPL, 2 wks - Dec 21]
      1. Testing and evaluation: PBS, LSF, etc.
      2. Installation of software
   4. MDSplus Software [MIT, 1 wk - date depends on Globus]
      1. Installed on cluster
      2. Interface MDSplus and Globus on cluster
      3. Data connection to MDSplus Servers at C-Mod, DIII-D, NSTX
   5. Relational Database Software [PPPL, 1 wk - date depends on above task]
      1. Client installed on cluster
      2. Data connection to SQL Servers and Run Management DB at C-Mod and DIII-D
2. Modification of PPPL Linux TRANSP to read from and write to MDSplus directly [PPPL, 4 wks - date depends on MDSplus software task above]
   1. Coordinate with DIII-D and C-Mod
3. Modification of DIII-D MDSplus TRANSP Trees
   1. Storage of output data [GA, 1 wk - Dec 21]
      1. Modified to be consistent with C-Mod trees
      2. Load a representative sample of DIII-D TRANSP runs
   2. Storage of input data [GA, 4 wks - Jan 31]
      1. Design input tree structure - consider C-Mod [GA - Dec 21]
      2. Implementation of input tree structure [GA - Jan 31]
4. GUI for TRANSP Runs [GA, 3 wks - Feb 15]
   1. Design of GUI [GA - Jan 18]
      1. General so that all labs can use
      2. Decide on a language
   2. Implementation of GUI [GA - Jan 31]
      1. Create general structure
      2. Hook in general data
      3. Hook in profile data
      4. Hook in auxiliary heating data
   3. Test GUI at multiple sites [GA - Feb 15]
      1. Loading data into MDSplus trees
      2. remotely starting TRANSP at PPPL
      3. Modify GUI as required
5. IDL-based Visualization of TRANSP Data [MIT, 2 wks - Jan 15]
   1. Modification of MultiGraph
      1. Read from multiple MDSplus Servers
      2. Read multigraph data from MDSplus
6. Monitoring of TRANSP runs
   1. Detailed specification listing for job monitoring[ANL, ? wks - Dec 21]
      1. Document with list of capabilities and interface including required level of access control
   2. Details of present Globus capabilities[ANL, ? wks - Jan 10]
      1. Specification of required Globus modifications with timeline
   3. Modifications to Globus
      1. Specified after Jan 10 Workshop at MIT
7. Use Policy
   1. Simple yes or no policy for the Linux CPU cycles
      1. Akenti called from either the gatekeeper or the job manager [LBNL, 3 wks, Jan 10 - Feb 1]
      2. Done from the Globus installation on Linux Cluster [LBNL,PPPL, 1 wk Feb 8]
   2. Will not address qualified yes answers at this time
   3. Akenti/MDSplus [MIT, 3 wks - Feb 11 - Mar 1 ]
      1. Method One: Akenti with user account mapping
      2. Method Two: Tree access policies directly to Akenti
      3. Method One will be investigated initially
      4. Install Akenti on MDSplus servers [LBNL, 1 wk - Dec 21]
      5. Establish what are the resources at each site are and set up a root policy for each one. [LBNL, 1 wk - Jan 10]
      6. Investigate having GRAM call out to Akenti for authorization information instead of just using the grid-map file. [LBNL&ANL, 1 wk - Feb 1]
8. CA For Fusion Collaboratory
   1. LBNL will be the CA for Fusion Collabortory
      1. Mary can be the RA as we need to add people
      2. As the DOE CA comes up this will fade away
   2. DOE ESNET CA - Interim Certificates by Dec 15, permanent by Jan 15
      1. Mary interface to this group - we are the customer
      2. Initially Mary acts as RA
      3. When this is active, GA, MIT, PPPL will act as an RA
      4. Mary to Instruct Fusion People how to create Certificates [ LBNL Jan 10]
9. Additional Globus capabilities to work with MDSplus
   1. Obstacles to Globus Deployment: C-Mod/NSTX are OpenVMS, DIII-D is Compaq True-64 Unix. [ANL, MIT, ? wks - Jan 10]
      1. Develop a design where Globus authentication and security mechanisms are used to obtain a connection to an entity acting as a gateway to an MDSplus server.
      2. Design a custom solution whereby the gateway is used to establish an authenticated and secure connection to the target platform.
      3. Linux server as a wrapper to OpenVMS & True-64 including Globus and Akenti. [MIT, 3 wks - ?]
   2. Obstacles to Globus Deployment: MDSplus uses inetd, Globus does not [ANL, ? wks - Jan 10]
      1. Globus has no provision for working with remote applications activated by inetd yet this is the mode of operation of MDSplus servers.
      2. Have a clear understanding of the needs of MDSplus and options available in Globus for dealing with this situation.
      3. Evaluate, select, and implement the best option.
10. Secure connection to Microsoft SQL Database (Run Management)
    1. Experiment with Java CoG interface
       1. can this replace Globus on Windows?
    2. Globus port to windows?
       1. Complete port is estimated for Summer 2002
       2. Port of common and GSI modules maybe by March 2002
       3. If this works need globus version of MDSplus client/server on Windows. [MIT, 2 wks - Date depends on Globus port]
    3. Ian - Information from your Microsoft contact?
11. Coordinate demonstration with local arrangements at Sherwood and TTF [GA, 1 wk - March 31]
    1. Sherwood with Riccardo Betti from the University of Rochester
       1. Issues include Internet access, power, furniture, workstations
    2. TTF with Ed Synakowski from PPPL
       1. Issues include Internet access, power, furniture, workstations

Implementation Philosophy Focused on November 2002
Demonstration - November 2002
Working Plan for November 2002

1. What is our focus here?

team contact info | about the fusion grid | fusiongrid research Last modified 12/11/01. Comments? webmaster