Proposed Questions for the AY 2008 NERSC Allocations Request Form

Project Overview

1. Principal Investigator
2. Other Authorized Preparers
3. Senior Investigators
4. Project Title
5. Project Name
6. Project Class (DOE Base, SciDAC, Startup, Guests, Overhead)
7. Sponsoring Site
8. Science Category
9. DOE Office and Program

DOE Funding

10. Is this project funded by the DOE Office of Science?
11. Does this project make use of national security information? (yes | no)

    Does this project collect and store Social Security Numbers, Personally Identifiable Heatlh Information (names or other identifiers matched with health data), Driver's License Numbers, or Financial Account Numbers? (yes | no)

    If this project is given a NERSC award ... (keep these two questions)

Resources Requested

12. Computational Resources Requested

                                           Hours       Hours
    Center   Resource  Alloc Type   Repo   Used 2007   Requested 2008  
    
    

13. Mass Storage Resources Requested

                                           SRUs        SRUs
    Center   Resource  Alloc Type   Repo   Used 2007   Requested 2008 
    
    NERSC    HPSS
    

14. Justification for Resources Requested

PDSF

15. PDSF Repos

    [existing PDSF repos renewed here]

    Do you need a new PDSF project account? Y/N
    If yes, please enter your experiment name (3-8 characters):

Project Description

16. Project Description
    Describe the research you are performing or will perform with NERSC resources. Your responses may be included in the NERSC Annual Report. Please update for the next year. (4000 character maximum.)
    • 16.1 Project Summary: Provide a brief project description. What will this project accomplish? What is the significance of this work?
    • 16.2 URL for a relevant web page. Enter ONLY ONE using the format http://www.nersc.gov
      Can we figure out a way to allow multiple entries and still display them as links?
17. Accomplishments
    Describe accomplishments made on this project in the previous year. State how the use of NERSC resources allowed you to meet your research objectives. Your response may be included in the NERSC Annual Report. (4000 character maximum.)
18. Relevant Publications: enter in the order authors, title, journal

Can we collect this in a systematic way such that it is easy to produce the pubs list?
• 18.1 Refereed Publications: List all refereed publications in the last 12 months based on research using NERSC resources. You may include publications submitted to journals but not publications in preparation.
• 18.2 Other Publications: List up to 5 other relevant publications in the last 12 months, also based on research using NERSC resources.

Code Descriptions

• Code Name
• Code Description (optional for application packages)
  • What does the code do?
  • What is its underlying mathematical formulation?
  • What are the numerical techniques employed?
  Provide 3 separate text boxes for each of the questions above.
• Machines to be used
  Where do you plan to run this code? What percent of the project's overall allocation do you expect this code to use on the machines you list?
• How many processors (or range of processors) do you plan to use for typical production runs?
• How do you decide on this number? Which of the factors below influence your choice:
  • computational requirements (grid points, number of particles, etc.)
  • memory required
  • queue policies, setup and turnaround times
  • need to run a very large number of serial (or small number of processor) jobs to process large amounts of data
  • other: ____________________________________
[NOTE: implement above as a checklist.]
• Code Algorithms (optional for application packages) [see checklist]
• Languages and Parallelization Methods Used (optional for application packages) [see checklist]
• Numerical and Performance Libraries Called (optional for application packages) [see checklist]
• Limits to Performance [see checklist]
  List the factors that limit code performance or scaling.
• Performance Comments
  Comments about the factors that limit performance.
• Does the code checkpoint? [Y/N]
• REMOVE THIS QUESTION -- Other Software Used
  Describe other software used in conjunction with this code: applications for pre and post processing, visualization, job control, etc.
• Plans for code development or enhancement

19.2 Code and Application Performance
Provide code performance data for typical processor counts used in production this past year. For machines with more than one processor per node enter # of procs as the number of nodes used times the number of processors per node.

code              # of                 Memory    how info was
name   machine    procs    Gflop/sec   (Gbytes)  collected / comments

____   _______    _____    ______      _______   ____________________
____   _______    _____    ______      _______   ____________________

Data / Analytics / Networking

20. Storage and Data Movement
    Describe your storage and data movement requirements:
    • What NERSC data storage facilities do you use or would you like to use? (NGF, HPSS, local scratch storage, etc.)
    • Between which systems (NERSC systems as well as remote systems) do you move data and how do you do it (or how would you like to do it)? (htar, hsi, pftp, gridFTP, secure file transfer, SRM, etc.)
    • How much space do you need:
    • Gigabytes of home (permanent) space
    • Total gigabytes/terabytes of scratch space
    • Gigabytes/Terabytes of scratch space needed for the largest job
    • Gigabytes/Terabtyes of project space in the NERSC Global Filesystem
    • Gigabytes/Terabytes in HPSS
21. Data and I/O
    • How much data does your project typically output in the course of the year?
    • What are your data characteristics?
      Please characterize the resolution and complexity of data produced by your simulation or collected and stored by your experiment. Describe characteristics such as:
      • Number of grid cells or zones
      • Typical number of variables per grid cell or data point
      • Grid dimensionality: 2D, 3D, 4D, higher
      • Grid meshing techniques used:

          __ no mesh                         
           __ structured rectilinear mesh             
           __ structured curvilinear mesh             
           __ unstructured mesh          
           __ hierarchical / adaptive mesh refinement  
           __ hybrid                        
        

        [NOTE: implement above as a checklist.]
    • How do you share data among your team members and how might you like to do it in the future if the present approach does not meet your needs? Does your project provide data to the public?
    • Do you do data subsetting? Do you (or would you like to) make use of any database systems or other similar technology for performing subset extraction?
    • How do you perform I/O? Do you write concurrently from all processors to shared files (parallel I/O) or one file per processor or node? Have you experienced problems or constraints due to I/O?
    • What file formats / data libraries do you use?
      • POSIX
      • F77 unformatted binary
      • MPI/IO
      • Parallel NetCDF
      • NetCDF
      • HDF5
      • H5Part
      • HDF4
      • FITS
      • NCARU
      • Global Arrays
      • community file format, describe: ______________________________
      • custom file format, describe: ______________________________
      • database format, describe: ______________________________
      • other, describe: ______________________________
      [NOTE: implement above as a checklist.]
    • REMOVE THIS QUESTION -- If you have a data intensive computing project (defined as a project which needs access to more than 500 Gigabytes of storage) please provide an overview of how your project produces, processes, manages and transfers data.
22. Analytics Applications
    Which applications/libraries do you use or intend to use for data analysis and visualization at NERSC?
    • AVS
    • AVS/Express
    • EnSight
    • ferret
    • Garlic
    • GIMP
    • Gnuplot
    • Grace
    • GrADS
    • Gsharp
    • IDL
    • ImageMagick
    • Maple
    • Mathematica
    • MATLAB
    • MySQL
    • NCAR
    • ParaView
    • PERL
    • pgplot
    • PostgreSQL
    • Python
    • R
    • Rasmol
    • ROOT
    • Tcl/Tk
    • VisIt
    • VMD
    • Other, describe: __________________________________
    [NOTE: implement above as a checklist.]
23. Networking
    If current network bandwidth between your site and NERSC does not meet your needs, please explain.

Other Info / Finish

24. Other HPC Support
    List High Performance Computing support from outside NERSC for this project you currently have or expect in the next year.
25. Additional Information
    Enter any additional information you would like the review committees to consider in support of this request.
26. Feedback and Project Requirements
    Use this space to provide feedback on the allocations process or make other comments. If your project has special requirements that you would like NERSC to consider, describe them here.
27. Current Request Status

algorithms checklist

Structured Mesh		Direct Solver
Unstructured Mesh	Dense Linear Algebra
Adaptive Mesh		Sparse Linear Algebra
No Mesh			Heavy use of integer/logical arithmetic
Iterative Solver	Monte Carlo
Particle Methods	FFT
Multigrid		Other: _____________________

Particle methods
  __ Particle In Cell (PIC)
  __ Particle-in-Mesh
  __ N-Body
  __ Fast Multipole
  __ Barnes Hut
  __ Other: _________________________________

 Linear equationa solvers                         
   __ dense solver 
   __ sparse direct solver                   
   __ iterative method 
   __ multigrids/multilevels                 FFT
                                                 __ Local (no inter-node
                                                    communication is required)
                                                 __ Global (which span multiple
                                                    processors and/or nodes)
Eigenvalue solvers
   __ dense eigensolver                  
   __ sparse eigensolver                  

Other methods
   __ Monte Carlo
   __ Genomic search
   __ Nonlinear equations solver 
   __ Ordinary differential equation solver
   __ Numerical optimization
   __ Quadrature rules
   __ Interpolation, spline approximation 
   __ Use of extended precision floating-point arithmetic
   __ Heavy use of integer/logical arithmetic
   __ Other algorithms: ________________________________



languages checklist
Assembler      C	    C++	    Fortran90	
Fortran95      Java     UPC       Python   
MPI            LAPI	    OpenMP    Other _____________




numerical and performance libraries checklist

Aztec     Chombo      DIMEMAS     ESSL        FFTW         HPMLIB    HYPRE
IMSL      LAPACK      MASS        METIS        NAG      
Overture  PAPI        PARAVER     PARPACK  
PESSL     PETSC       PWSMP       ScaLAPACK    SLEPc     SUNDIALS  superlu  
TAO       TAU         Trilinos    VAMPIR
WSMP      Other: ______________________




limits to performance checklist
Algorithm scaling	   	All-to-all performance (large messages)
Load imbalance	          All-reduce performance (small messages)
Small problem size	   	Broadcast performance
Point-to-point latency	 Barrier performance
Point-to-point bandwidth   	MPI I/O
Available disk space	 POSIX I/O
Bug			 External network
Unknown   		Single node memory performance
Other: ________________