Application of High-performance Visual Analysis Methods to Laser Wakefield Particle Acceleration Data Page: 1 of 2
This article is part of the collection entitled: Office of Scientific & Technical Information Technical Reports and was provided to Digital Library by the UNT Libraries Government Documents Department.
The following text was automatically extracted from the image on this page using optical character recognition software:
Application of High-performance Visual Analysis Methods to Laser
Wakefield Particle Acceleration Data
Oliver RObel **, Prabhat *, Kesheng Wu *, Hank Childs , Jeremy Meredith ,
Cameron G.R. Geddes , Estelle Cormier-Michel , Sean Ahern 1, Gunther H. Weber *,
Peter Messmer **, Hans Hagen t, Bernd Hamann ** and E. Wes Bethel **
Our work combines and extends techniques from high-performance
scientific data management and visualization to enable scientific
researchers to gain insight from extremely large, complex, time-
varying laser wakefield particle accelerator simulation data. We ex-
tend histogram-based parallel coordinates for use in visual informa-
tion display as well as an interface for guiding and performing data
mining operations, which are based upon multi-dimensional and
temporal thresholding and data subsetting operations. To achieve
very high performance on parallel computing platforms, we lever-
age FastBit, a state-of-the-art index/query technology, to acceler-
ate data mining and multi-dimensional histogram computation. We
show how these techniques are used in practice by scientific re-
searchers to identify, visualize and analyze a particle beam in a
large, time-varying dataset.
Index Terms: data mining, visual data analysis, accelerator mod-
eling, parallel visualization, large data visualization, temporal visu-
alization, temporal data analysis
Laser WakeField Accelerators (LWFAs) promise to be a compact
source of high-energy electron beams and radiation. Using an in-
tense laser pulse fired into a plasma, LWFAs have been shown to
generate electric fields thousands of times stronger than those in
conventional particle accelerators. LWFAs accelerate particles to
high energies of 1GeV within 3cm compared to >~ t 5m in tradi-
tional electromagnetic accelerators .
In order to gain deeper understanding of phenomena observed in
LWFA experiments, researchers at the LOASIS facility at LBNL 
and at Tech-X** use the VORPAL  simulation code to computa-
tionally model their experiments. Gaining scientific insight is often
challenging due to the size and complexity of the data output by the
simulation. To support scientific knowledge discovery and hypoth-
esis testing on this kind of data, we combine and extend two dif-
ferent technologies aimed at enabling rapid, interactive visual data
exploration and analysis. We extend and adopted histogram-based
parallel coordinates as a vehicle for visual information display and
as an interface for data selection via multi-dimensional threshold-
ing. We employ state-of-the-art index/query technology to accel-
erate the data mining process as well as generation of conditional
histograms. We execute these activities on large, parallel machines
*Computational Research Division, Lawrence Berkeley National Labo-
ratory (LBNL), One Cyclotron Road, Berkeley, CA 94720, USA.
tInternational Research Training Group "Visualization of Large and Un-
structured Data Sets," University of Kaiserslautern, Germany.
*Institute for Data Analysis and Visualization, University of California,
Davis, One Shields Avenue, Davis, CA 95616, USA.
Lawrence Livermore National Laboratory, 7000 East Avenue, Liver-
more, CA 94550
1oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, TN 37831,
IILOASIS program of Lawrence Berkeley National Laboratory, One Cy-
clotron Road, Berkeley, CA 94720, USA.
**Tech-X Corporation, 5621 Arapahoe Ave. Suite A, Boulder, CO 80303.
Figure 1: Histogram-based parallel coordinates using 32x32 uniform
binning (left) and adaptive binning (right). The adaptive binning pre-
serves more details in dense areas while discarding some details in
sparse areas of the data which may ease comparison of selections
(red) with density-features of the data in low-level-of-detail views.
to achieve interactive data mining and visual analysis rates on the
order of seconds for terabyte-size datasets.
Historically, a scientific researcher investigates time-varying
data of this type by first creating an animation, then viewing it to
identify a timestep and multivariate threshold values that will result
in isolating particles that form a beam. Having defined the beam,
one would run scripts searching at each timestep for particles of in-
terest to perform particle tracing. In the past, these activities have
typically required hours to complete. Using our techniques and im-
plementation, this activity is executed at interactive rates: a scien-
tific researcher can interactively select a beam and trace it within
seconds, greatly reducing the duty cycle in visual data exploration
and mining while improving accuracy of the analysis .
2 RELATED WORK
Parallel coordinates are a common information visualization tech-
nique for visual display of high-dimensional data sets. In the case
of extremely large data, standard, line-based parallel coordinates
suffer from problems like excessive clutter and occlusion. We have
built on and extend the work of Novotnj and Hauser  who binned
parallel coordinates to overcome these limitations, which are espe-
cially acute with very large datasets.
We utilize FastBit , a state-of-the-art index/query system for
data extraction and subsetting. It implements the fastest-known
bitmap compression technique, and has been demonstrated to be
effective in a number of data analysis applications. In particular, it
has a number of efficient functions for computing conditional his-
tograms, which are crucial for this work.
To deliver our research results to scientific researchers, our new
techniques are implemented in VisIt . VisIt is a production-
quality, parallel capable visual data analysis application that runs
on virtually all modern HPC platforms.
3 SYSTEM DESIGN
To accelerate data mining operations, we use FastBit, a state-of-
the-art data management technology for indexing and searching.
We use FastBit to perform data subsetting/selection and to compute
conditional histograms. We implemented these operations using
FastBit directly in the file-reader stage of the processing pipeline
in VisIt. The conditional histograms serve as basis for the visual
presentation of data vis-a-vis histogram-based parallel coordinates.
Here’s what’s next.
This article can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Article.
Rubel, Oliver; Prabhat, Mr.; Wu, Kesheng; Childs, Hank; Meredith, Jeremy; Geddes, Cameron G.R. et al. Application of High-performance Visual Analysis Methods to Laser Wakefield Particle Acceleration Data, article, August 28, 2008; Berkeley, California. (https://digital.library.unt.edu/ark:/67531/metadc895653/m1/1/: accessed March 19, 2019), University of North Texas Libraries, Digital Library, https://digital.library.unt.edu; crediting UNT Libraries Government Documents Department.