SCD News: May 27, 2005
Program includes presentations, tutorial, roundtable, posters, and reception
|
A small subset of participants in the CISL 2005 User Forum pose for a group photo during a conference break. More photos |
N CAR's Computational and Information Systems Laboratory (CISL) welcomed 115 registrants to its 2005 User Forum, held May 17-19, 2005, at the NCAR Mesa Lab in Boulder, Colorado. Participants from NCAR, CISL, and the university community had a chance to share ideas and establish new connections with one another over the course of three days.
The program included 30 presentations on topics such as:
Presentations are available online at http://www.scd.ucar.edu/info/UserForum/agenda.html .
The conference included a half-day tutorial on the Earth System Modeling Framework (ESMF), a lunchtime roundtable discussion, an afternoon poster session, and an evening reception. See the photo gallery
“The User Forum is now a regular event for CISL/SCD users,” says Tom Bettge, associate director of SCD Operations and Services. “It's a great opportunity for SCD staff to exchange ideas, plans, and requirements with the community. Emails and phone calls allow for communication, but nothing beats face-to-face interaction. And we've found that knowing how users manage their workloads and workflows is extremely helpful in planning to meet their needs.”
“We were excited by the turnout from both universities and NCAR divisions,” says conference organizer Ginger Caldwell, an SCD allocations management specialist. “We invited speakers from four universities and three NCAR divisions to share their experiences using SCD services. They provided a lot of interesting information and gave feedback on what resources they would like to see available from SCD in the future.” She notes that CISL is already making plans for a User Forum to be held in 2006.
“The User Forum is important because it's a chance for us to tell our customers what we're doing and for them to tell us what they want,” says Dick Valent, head of SCD's Consulting Services. “For example, we highlighted services we offer such as the new web interface to the Mass Storage System; we heard how important NCL was to the user community. During breaks and after sessions, we had plenty of opportunities to talk to users about specific concerns.”
The conference opened with an afternoon tutorial on the Earth System Modeling Framework (ESMF) given by Cecelia Deluca, ESMF technical project manager, and Nancy Collins, a software engineer in SCD's ESMF Group. ESMF is a software package for building scalable, portable, modular modeling and data assimilation applications.
The SCD Visualization Lab was filled to capacity, with 34 attendees. Cecelia and Nancy presented the goals and scope of the ESMF project, described the current status of its software implementation, and outlined the overall principles underlying the software. They described ESMF functionality to give users an understanding of how ESMF might be utilized in their own codes and described how codes incorporate and run under ESMF.
Pat
Behling (Center for Climatic Research, University of
Wisconsin-Madison) hosted a lunchtime roundtable to discuss strategies for providing
training to graduate students in the use of computational tools. Participants
offered a wide range of perspectives, noting that Fortran was no longer widely
taught in universities. Workshops, books, tutorials, and documentation were all
found to be helpful.
Vijay Kumar Tallapragada (Department of Meteorology, Florida State University), described how he and his colleagues at FSU used ERA-40 datasets from SCD in their work with numerical weather prediction, climate modeling, and observational studies. “We largely depend on the datasets from NCAR/SCD for these studies,” he said. “The support form SCD in terms of data distribution, storage, and computing cannot be overstated. The ERA-40 datasets—particularly the corrected U, V [horizontal winds] datasets from SCD—have provided us with more choices and added confidence to our efforts. We anticipate more aggressive use of these datasets in many of our ongoing projects.”
Wei
Wang (NCAR's Mesoscale and Microscale Meteorology Division) talked
about how MMM researchers used bluesky, the IBM Cluster 1600 supported by
SCD, to run the Weather Research and Forecasting Model (WRF) for a series
of near-real-time high-resolution forecast experiments. These included the
Bow Echo and Mesoscale Convective Vortex Experiment (BAMEX) in 2003; projects
conducted in spring and summer of 2004 and 2005; and hurricane studies in
2003 and 2004. Wei and her colleagues used 128 to 256 processors of bluesky
for up to six wallclock hours per day, storing up to 70 gigabytes of data
per day on the MSS. She noted that experiments were of tremendous benefit
to the forecasting community and thanked SCD for working with her team over
the past two years to make the forecasts successful.
Gary
Strand (NCAR Climate and Global Dynamics Division) discussed NCAR's
contribution to the Intergovernmental Panel on Climate Change (IPCC) 4th
Assessment Report. The IPCC project was the first large-scale coordination
of modeling efforts, data analysis, data management, and data dissemination
by the global climate modeling community. CGD, which provided output from
the Parallel Climate Model 1 and the Community Climate System Model 3, made
use of tempest (an SCD-supported data analysis supercomputer) for CPU-intensive
data analysis and processing, and published about 20 terabytes of data on
SCD's Earth System Grid portal. Gary noted that hardware resources have been
inadequate and offered several suggestions to improve data management for
climate simulations.
Bjorn
Stevens (University of California Los Angeles) noted that SCD success
is often equated with flops, but that its other resources—people, software,
the Mass Storage System—are equally essential. From the university perspective,
he said, SCD/CISL are “tool builders”; what makes the division so important
is that the tools it provides make it easier for users to do what they want
to do. He considers the NCAR Command Language (NCL) a “spectacular” tool
for file handling and analysis.
Tom
Bettge gave
an update on SCD, reviewing recent highlights and previewing what
to expect in years to come. As always, he said, science is driving the demand
for scientific computing, and the demand for compute cycles exceeds availability.
SCD supercomputers are well utilized, and computational resources have been
made available for numerous field projects. Data storage in the MSS continues
to grow at an exponential rate: it took 18 years to reach one petabyte of stored
data, and only 18 months to reach the second petabyte. SCD is working to achieve
a 25-fold increase over current sustained computing capacity in the next five
years. However, because the Mesa Lab Computer Room will reach power, cooling,
and space limits with the 2006 computing acquisition, SCD is looking
into options for expanding the data center.
Tom emphasized that SCD remains committed to a balanced and sustained investment in a robust cyberinfrastructure, including supercomputing systems, mass storage, networking, data management, software tools and frameworks, software and expertise, and security.
Aaron
Anderson, manager of SCD's Enterprise Services Section, spoke about
the need for an SCD computing center expansion. Science determines computational
requirements, he said, and model complexity is increasing. Computer chips
are faster and cheaper—but hotter. The NCAR Computer Room, built in 1976
in an era of low power requirements, is obsolescent. If nothing is done,
the power and design limitations of the infrastructure will become a serious
constraint on NCAR's scientific computing capacity. After looking
at a number of options, SCD has decided to maximize the life and capabilities
of the Mesa Lab facility, but move forward on building a data center expansion.
SCD is currently completing conceptual design and cost estimates and developing
funding strategies.
Rich
Loft, SCD deputy director, gave an overview of NCAR's new Blue
Gene/L computer, which arrived on March 15, 2005. Blue Gene/L is a massively
parallel system with 2,048 processors, a peak speed of 5.73 teraflops, and
high efficiency in terms of gigaflops per kilowatt. NCAR's Blue Gene/L runs
the Linpack benchmark at 4.62 teraflops and is unofficially the 33rd fastest
system in the world—in
just one rack. Because the machine is experimental and lacks a complete computational
environment, it will be used to explore and test this new computer architecture.
SCD has configured the machine, solved various system problems, and is now
working with other BG/L sites and the BG/L Consortium to develop scientific
and technical applications for the system.
Rich Johnson, a security engineer in SCD's Enterprise Services Section (ESS), gave an overview of a series of security incidents in 2004 that affected NCAR and a number of high-performance computing centers, universities, and national laboratories across the U.S. and Europe. Daryl Herzmann (Department of Agronomy, Iowa State University) described the impact of the incidents from the perspective of a university user. Lynda McGinley, an ESS security engineer, talked about the importance of security awareness. She provided a list of tools and techniques for users and system administrators that would help protect workstations and laptops.
Steven
Worley, manager of SCD's Data Support Section (DSS), and DSS software
engineers Joey Comeaux, Chi-Fan Shih, Doug Schuster, and Bob
Dattore described
new features in SCD's research data archive, a collection of more than 600
datasets used in the atmospheric and related sciences. Data is available
in many categories, including observations, model output, climatology data,
station summary data, satellite derived datasets, and global topology. DSS
offers data to users through different channels, including the web, FTP,
and the Community Data Portal. Users can now access ERA-40 reanalysis data,
North American Regional Reanalysis Data, and near real-time data from Unidata's
Internet Distribution (IDD) service. DSS is also working to develop digital
metadata to help with data discovery and access.
Cecelia Deluca talked about capabilities of the Earth System Modeling Framework. ESMF is a software package that provides tools for turning model codes into components with standard interfaces and standard drivers. It increases scientific productivity by making components easier to build, combine, and exchange, and by enabling modelers to take full advantage of high-end computers. It also promotes new scientific opportunities and services through community building and increased operability of codes.
Luca
Cinquini , a software engineer in SCD's Visualization and Enabling
Technologies Section, talked about two synergistic projects underway in SCD
aimed at advancing UCAR's cyberinfrastructure. The Community Data Portal,
a gateway to all data produced and stored at NCAR, promotes collaboration
and co-development among many research efforts within NCAR. The Earth System
Grid integrates supercomputers with large-scale data and analysis servers
located at numerous national labs and research centers to create a powerful
environment for next-generation climate research.
Pam
Gilman, head of SCD's Data Analysis Services Group (DASG), and John
Clyne, a DASG software engineer, spoke about SCD efforts to provide
a comprehensive environment for interactive data post-processing, analysis,
and visualization. Aiming to help users with scientific work flow, DASG supports
processing and visualization engines, high-performance storage, specialized
software, and the SCD Visualization Laboratory. The group is conducting basic
research and algorithm development of visualization and analysis methods.
John
Clyne of SCD's Data Analysis Services Group and Mark
Rast of NCAR's High Altitude Observatory discussed the challenges
associated with analyzing terascale turbulence datasets. John noted that
computational datasets are getting too big to analyze using traditional methods,
a problem that is especially acute in turbulence research. He discussed a
new software being developed in SCD, in collaboration with researchers from
the University of California-Davis and Ohio State, that employs new methods
for analyzing turbulence data. VAPoR, which stands for the Visualization
and Analysis Platform for ocean, atmosphere and sun Researchers, combines
a state-of-the-art visualizer with a state-of-the-art analyzer. Mark showed
several examples of how VAPoR could be used to explore complicated turbulence
datasets.
Lana Stillwell, head of SCD's Applications and Information
Services Group, spoke about the SCD Portal, an online gateway to SCD supercomputers
and account information. The portal provides an easy-to-use web interface to
SCD resources and information, allowing scientists to concentrate on science
and not computing. It provides current job queue and GAU information and is
customizable to provide only relevant information.
Mary
Haley, a software engineer in SCD's Visualization and Enabling Technologies
Section, introduced the audience to NCL and PyNGL, two programming languages
developed by SCD for the access, analysis, and visualization of data in the
geosciences. NCL, which stands for NCAR Command Language, is an interpreted
language that has powerful capabilities in file I/O, data analysis, and visualization.
Mary gave an overview of some of the strengths of NCL and discussed its new
features, including command-line options, contouring on triangular meshes,
enhancements to the GRIB reader, and a new website. PyNGL, which stands for
the Python interface to the NCL Graphics library, has generated an enthusiastic
response among atmospheric researchers who use Python for programming but
want the superior visualization capabilities of NCL. SCD has plans for improvements
and new capabilities in both NCL and PyNGL.
Dick
Valent, head of SCD's Consulting Services, described SCD's commitment to provide
customer assistance, training, help with flagship codes, and documentation
to users of SCD's high-end computational environment. He described the new
centralized service desk, which now offers user support 24x7. Many calls can
be handled immediately, day or night; others are routed to the right SCD staff.
Software engineers Mike Page, Juli Rew, and Siddhartha Ghosh gave targeted
presentations on how to best utilize SCD computational resources. Mike gave
tips on using the MSS efficiently and introduced a web interface for accessing
information on MSS holdings: http://www.scd.ucar.edu/docs/mss/MSSholdings.html.
Juli offered hints on IBM job scheduler and strategies for code porting and
efficiency. Sid gave advice on how to make secure file transfers to and from
the SCD supercomputers.
Al
Kellie, associate director of CISL, and Doug
Nychka, director of CISL's Institute for Mathematics Applied to
the Geosciences (IMAGe), stressed the institute's commitment to bring
mathematical tools to bear on fundamental problems in the geosciences.
Jeffrey
Anderson of the IMAGe Data Assimilation Research Section (DAReS)
gave a talk on the Data Assimilation Research Testbed (DART). DART is a data
assimilation facility that combines assimilation algorithms, models, and
observation sets. The DART software provides a flexible, extensible framework
for conducting data assimilation research on a wide variety of models and
observations. Many climate models are now DART-compliant, including the Weather
Research and Forecasting Model, the Community Atmosphere Model 2 and 3, ROSE,
and models from the Massachussets Institute of Technology, the Geophysical
Fluid Dynamics Laboratory, and the National Centers for Environmental Prediction.
Duane Rosenberg of the IMAGe Geophysical Turbulence Section spoke about his work with GASpAR, which stands for Geophysics/Astrophysics Spectral-element Adaptive Refinement. GASpAR is software designed to simulate geophysical fluid dynamics. Object-oriented and parallelized, it uses adaptive mesh refinement to solve linear and nonlinear advection problems. Duane showed various applications of GAsPAR and briefly outlined plans for future enhancements.
During the Wednesday afternoon break, participants viewed posters on display
on the mezzanine of the NCAR Mesa Lab. Steve Worley and Joey Comeau shared
information on ECMWF 45-year reanalysis data available from SCD. Amick St.-Cyr
and Stephen Thomas of SCD, Andrzej Wyszogrodski of IBM, and Wojciech Grabowski
of NCAR's Mesoscale and Microscale Meteorology Division were on hand to discuss
the coupling of sub-grid scale moist convection with global dynamics. Henry
Tufo, Matthew Woitaszek, and Jason Cope of SCD talked about a grid-enabled
terrestrial carbon-cycle modeling system called Grid-BGC. SCD's Michael Burek
provided expertise on SCD's Community Data Portal.
Speaker presentations given at the CISL 2005 User Forum are on the web at http://www.scd.ucar.edu/info/UserForum/agenda.html .
Full photo coverage is available at http://www.scd.ucar.edu/news/05/lead/userforum/index.html .
—Lynda Lester
Photos: Lynda Lester, NCAR/CISL
|
NCAR is managed
by UCAR and sponsored by the National Science Foundation |
