Docstoc

rast (PowerPoint)

Document Sample
rast (PowerPoint) Powered By Docstoc
					                                                       Kiepenheuer-Institut für Sonnenphysik
                                                       14 June 2006

VAPoR (Visualization and Analysis Platform for Ocean, Atmosphere, and Solar Researchers):
         Interactive analysis and visualization of very large data volumes

  Mark Rast
  Laboratory for Atmospheric and Space Physics
  Department of Astrophysical and Planetary Sciences
  University of Colorado, Boulder

  John Clyne and Alan Norton
  Scientific Computing Division
  National Center for Atmospheric Research
  Boulder, Colorado


       http://www.vapor.ucar.edu/
       • Freely available with support. Input into future capabilities.
Numerical models which can currently be run on typical supercomputing
platforms produce data in amounts that make storage expensive,
movement cumbersome, visualization difficult, and detailed analysis
impossible. The result is a significantly reduced scientific return from the
largest computational efforts.
1.              We can now compute more data than we can analyze.
                                  Performance gains from 1980 to present
              100000
                                                                                                               •   Not all technologies
              10000                                                                 Disk Drive Internal Data
                                                                                    Rate
                                                                                                                   advance at the same
                                                                                    Disk Drive Interface
                                                                                    Data Rate
                                                                                                                   rate
Improvement




               1000                                                                 Ethernet Network
                                                                                    Bandwidth
                                                                                                               •   Multiprocessor simulation
                                                                                    Intel Microprocessor
                                                                                    Clock Speed
                                                                                                                   vs. single/dual processor
                100
                                                                                    Drive Capacity                 analysis
                 10


                  1
                   80

                        82

                             84

                                  86

                                       88

                                            90

                                                 92

                                                      94

                                                           96

                                                                98

                                                                     00

                                                                          02

                                                                               04
                  19

                       19

                            19

                                 19

                                      19

                                           19

                                                19

                                                     19

                                                          19

                                                               19

                                                                    20

                                                                         20

                                                                              20




2.              Most analysis tools have poor volume visualization capabilities and
                most visualization tools have only rudimentary analysis capabilities.
Example: Compressible plume dynamics


• 504x504x2048
• 5 variables (u,v,w,rho,temp)
• ~500 time steps saved
• 9TBs storage
  (4GBs/variable/timestep)
• Six months compute time              QuickTime™ and a
                                         decompressor
                                 are neede d to see this picture.
  required on 112 IBM SP
  RS/6000 processors
 What is meant by interactive analysis?
  Definition: A system is interactive if the time between a user event and the
              response to that event is short enough maintain my full attention

  If the response time is…
        1-5 seconds : I’m engaged
        5-60 seconds : I’m tapping my foot
        1-3 minutes : I’m reading email
        > 3 minutes : I’ve forgotten why I asked the question!

                                                 Resolution   MBs per    Scalar      Vector
IO wait times for high resolution simulations:                variable   variable    variable
              •   Assumptions                                            wait time   wait time
                   – Single precision
                                                 1283         8          0.1         0.3
                   – 100 MB/sec bandwidth
                   – No contention               2563         67         0.7         2.1
                                                 5123         537        5.0         15.0
                                                 10243        4295       43.0        130.0

  Develop a tool with which one can interactively analyze
  and visualize very large data volumes.
                                     Rendering timings
                        10                                                                 1000




                                                                         Time in seconds
     Time in seconds




                                                                                           100
                         1

                                                                                             10                              Mdb
                                                         Mdb
                                                                                                                             Vtk
                        0.1
                                                                                              1



                       0.01                                                                 0.1
                              Full   1/2     1/4   1/8   Interactive                              Full   1/2     1/4   1/8

                                     Resolution                                                          Resolution


                       5123 Compressible Convection                             5042x2048 Compressible Plume
                                                         SGI Octane2,
                                                         1x600MHz R14k

                                                         SGI Origin,
                                                         10x600MHz R14k


Reduced resolution affords responsive interaction while preserving all but finest features.
                                     Calculation timings
                                   10000


                                   1000
                 Time in Seconds



                                    100
                                                                         pressure (eq 1)
                                     10                                  ionization (eq 2)
                                                                         enstrophy (eq 3)
                                      1

                                                                          SGI Origin,
                                     0.1                                  10x600MHz R14k

                                    0.01                                 Note: 1/2th resolution is
                                           Full   1/2        1/4   1/8   1/8th problem size, etc
                                                    Resolution
                          Interactive
5123 Compressible Convection

         Deriving new quantities on interactive time scales only
         possible with data reduction
Key VAPoR components:
      Multiresolution data access and subregion sampling
           Enable speed/quality tradeoffs
      Tightly coupled to existing analysis tools
           IDL, MatLab
      Advanced volume visualization tool
           Histogram based transfer funtion editor, Field line
           tracing, etc.


An interactive
multiresolution
visualization and
analysis tool.
Wavelet Transforms for 3D Multiresolution data representation:

 • Hierarchical data representation
 • Invertible and lossless (subject
   to floating point round off errors)
 • Numerically efficient
 • No additional storage cost

Example: Haar Wavelet
(current VAPoR format)

            P( x)  x
Haar
                    1
operators   U ( x)  x
                    2


                                               Store averages and
                                               differences.
Compressible Convection




                         Rast, 2002
   1283           5123
Compressible
   plume




   Compressible plume
   data set shown at
   native and
   progressively coarser
   resolutions


                                                                                    Rast, 2002

              Resolution:   504x504x2048   252x252x1024   126x126x512   63x63x256
            Problem size:       Full           1/8            1/64        1/512
A test of multiresolution analysis:
Force balance in supersonic downflows
                                                Resolution
                                          p                            u                           p
    u                                   g                                                           g
                                          z                                                             z
                                                           Full
      u2                             1 p                                  u2                      1 p
       r                               r          Half                     r                        r

                                                           Subdomain
                                                 z
                                                           selection and                                        z
                                                           reduced
                                                    u2    resolution                                              u2
                                                     r     together yield                                           r

                                                    1 p   data reduction                                          1 p
                                                     r                                                            r
                                                           by a factor of
                                                           128!
             Sites of supersonic downflow are also those of very high vertical vorticity. The cores of the vortex
             tubes are evacuated, with centripetal acceleration balancing that due to the inward directed pressure
             gradient. Buoyancy forces are maximum on the tube periphery due to mass flux convergence.
             The same interpretation results from analysis at half resolution.
Future Plans:

   • Incorporate visualization techniques based on scientists’ needs
      – Nonuniform grids
      – Adaptive grids
   • Understand effect of data compression
      – Error analysis and error visualization
      – Obtain bounds on degradation of analysis results
   • Explore lossy data compression
   • Improve access to terabyte datasets
      – Multiresolution data output as a byproduct of the simulation

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:13
posted:3/17/2012
language:
pages:12