NCSA NEWS

Making science happen

Story posted May 9, 2007

By Barbara Jewett, NCSA

The user support offered by NCSA is unsurpassed. From answering the simplest question to assisting with complicated codes, advanced visualizations, or dedicated runs, NCSA has developed a level of user support that keeps science moving forward.

Microphysical processes in tropical cyclones. Thermal conduction in nanofluids. Formation of galaxies and supermassive black holes. Risk assessment of aerosol dispersion in a city. Determining pavement damage owing to different axle configurations. These are just a few of the research topics investigated using NCSA compute resources last year. In fact, nearly 1,400 scientists, engineers, and students racked up more than 717 million normalized units of computing time. And they once again gave NCSA top marks for user support services in the 2006 Cyberinfrastructure Partnership user survey because they get quick responses to their simple questions and the technical assistance they require for the more advanced ones.

Support for projects big or small

Support at NCSA takes many paths. Here's a brief look at just a few of the many ways researchers are assisted, whether their project is running solely on NCSA machines or involves TeraGrid resources.

The first route is the help desk, which provides phone and email support around the clock, every day of the year. For slightly more difficult issues, the consulting office is the next source of support. The consulting office aids all users of NCSA resources and services. They help users learn how to run their jobs on NCSA machines, including running batch jobs, or providing some programming assistance, either in general or for a particular application. The fields of computational chemistry, computational biology, structural mechanics, computational fluid dynamics, math, and physics require specialized expertise; the consulting office staff includes people proficient in the software programs these researchers use with the know-how to capably handle other issues.

"The NCSA consulting office was extremely helpful to me," says Renyue Cen, a senior research astronomer at Princeton University. "They implemented a required faster, parallel I/O scheme into the TIGER code and helped debug the code. As a result, we were able to make the world's largest cosmological hydrodynamic simulation with 8 billion cells."

Aside from the scientific results Cen obtained, this simulation was visualized by NCSA's Advanced Visualization Laboratory and featured in the PBS NOVA program "Monster of the Milky Way" that aired last October.

Getting researchers past all the barriers and making science happen is the objective of NCSA, and nowhere is that more evident than in the area of advanced support. Providing the missing link in a research project sometimes requires that user support experts focus on a particular researcher's stumbling block to success. To move the research forward might mean creating custom software or benchmarking systems. Dedicated services range from short-term efforts in understanding code on a particular processor to months-long dealings to achieve goals in areas such as scalability, large dataset management, and advanced visualization. Current projects include assisting with simulating novel biofuel liquid combustors, developing computational methods for large-scale simulation of buildings subjected to earthquakes and benchmarking systems at NCSA and the San Diego Supercomputer Center (SDSC).

Capability computing is a key component to helping researchers achieve success. For example, David Baker of the University of Washington developed a code, called Rosetta, to model the structure of proteins. Baker's team worked with SDSC to adapt the code to run on high-performance computing systems, then turned to NCSA for compute power. It was quickly apparent that in order for Baker to make progress with his research he would need substantial fractions of available resources, so NCSA set aside 128 nodes of the Tungsten cluster for several months for his use.

Baker also developed a Web portal, called Robetta, which allows other researchers to run Rosetta jobs. Initially, Robetta jobs were run on a small cluster in Baker's lab, but when submissions exceeded capacity, NCSA was able to provide a solution. Working with Baker's team and the Condor development team from the University of Wisconsin, NCSA was able to integrate the Condor workload management system supporting Robetta with the center's Condor cluster, effectively mirroring the Robetta server at NCSA.

But Baker is not alone in his need for dedicated resources. At any given time last year, up to 60 percent of Tungsten's capacity was in use for dedicated allocations. And the requests for dedicated allocations have risen so dramatically that NCSA has installed a new almost 90 teraflop system, Abe, to meet the need. The addition of Abe also allows the center to further expand our capability computing responsiveness, whether it is dedicated allocations or on-demand scheduling for running simulations, such as in disaster situations.

We have data...now what?

Large project or small, all research utilizing NCSA has some basic parameters regardless of subject matter: Researchers generate data, next they need to process it, move it, and store it.

Answering the questions of how to deal with data -- which, as in the case of astrophysics data from telescopes can be in the realm of tens of terabytes -- is another area in which NCSA supports scientists and engineers. Users at NCSA have access to high-speed parallel file systems on each of our platforms to support the creation and analysis of large data sets. There is more than a petabyte of raw disk capacity distributed among our machines as well as a long-term data storage system with several petabytes of archival capacity.

For instance, the Large Synoptic Survey Telescope (LSST) project will generate an estimated 15 terabytes of raw data and 100 terabytes of processed data every night when it begins operating in 2013. To lay the foundation for the LSST cyberenvironment, NCSA and the National Optical Astronomy Observatory (NOAO) are developing a prototype data pipeline using the vast stores of data generated by the ground-based observatories NOAO oversees. NCSA created a mirror of the NOAO data archive in Urbana. The archive replication system is built on the Storage Resource Broker middleware developed at SDSC and a transfer queuing system NCSA developed for its archive of data from the Berkeley-Illinois-Maryland Array radio telescope.

Mirroring NOAO's archive not only gives astronomers a high-bandwidth site from which to access data, it is also part of a strategy of security through redundancy, ensuring that data will survive at one site even if a catastrophic event hits another site. NCSA's robust mass storage system archives more than 3 petabytes of researchers' data, with data added at a rate of 40 to 60 terabytes each month. However, the data management expertise of the center's staff -- which encompasses experience running large file systems, parallel file systems, many storage architecture types, HPC storage, and database storage -- is as much of an asset as the storage infrastructure.

As challenging as the tasks of capturing, moving, processing, and storing data are, they are just the preliminary steps. The real excitement begins when researchers can access and analyze the data. The center provides software to manipulate the data, and NCSA experts assist as needed with data management and visualizations.

For LSST, community access will be provided through a Web-based virtual observatory (VO). NOAO and NCSA are working to develop the VO model and the VO tools, including an authentication and authorization framework for the NOAO portal, an online tool that enables users to find, access, and analyze the data available through multiple public archives, such as the Sloan Digital Sky Survey, Canadian Network for Observational Cosmology, Chandra X-ray Observatory Center, and others.