UW’s Wang Focuses on Optimizing Supercomputer, Improving Reliability of High-Performance Computers
June 6, 2012 — Numerous University of Wyoming professors plan to use the NCAR-Wyoming Supercomputing Center (NWSC) for their high-level computational research this fall. Liqiang Wang is the person who is going to help make sure things go off without a hitch.
In a case of personal research matching the needs of the National Center for Atmospheric Research, Wang, a UW associate professor of computer science, will help optimize use of the supercomputer in Cheyenne and improve the reliability of HPC (high-performance computing) programs.
"We want all codes that run on Yellowstone (the supercomputer's name) to be more scalable, more reliable and perform better," Wang says. "We want to make the process easier and faster on the supercomputer. We'll be making it easier for other researchers to do their research."
For example, a researcher's code may run correctly on a small computer cluster. However, it may not be able to scale up -- without adjustments -- for use on a supercomputer, Wang says.
"Worse, the code may crash. It's happened a lot in my own research projects," he says.
Fine-tuning the nuts and bolts
Supported by the National Science Foundation, including the NSF Career Award, Wang's research focuses on the analysis and design of HPC systems. For analysis, Wang works to detect programming errors and performance defects for large-scale HPC programs. For design, he is investigating data-intensive parallel computing on multi-core central processing units (CPU), graphics processing units (GPU) and cloud computing platforms.
Wang is collaborating with NCAR researchers to trouble-shoot and scale up several widely used HPC codes.
What initially piqued Wang's interest in trouble-shooting computer code? During his undergraduate days in China, and while working on his doctorate at Stony Brook, Wang wrote numerous computer code. Inevitably, there would be problems.
"It was a big headache for me. It took me a long time to debug code," he recalls, shaking his head. "It motivated me to do some research, to find the problems automatically instead of manually."
"When debugging code, people usually detect programming errors manually," Wang continues. "We try to solve problems automatically, so they can just click a button and see what the errors are."
The buddy system
In addition to assisting NCAR, Wang is collaborating with Po Chen, a UW associate professor of geology and geophysics. They will use the supercomputer to create more detailed images of his computer models of underground structures to better predict seismic activity. This includes earthquakes in Southern California and illegal underground nuclear explosions in Eastern Eurasia.
Wang and his students are designing a more scalable code called PLSQR code, which can compute the data more efficiently for seismic events Po is researching. PLSQR code is a scalable and parallel solver for linear equations, and can be applied in the computation of large-scale data sets.
"In seismic tomography, which creates images of the underground earth structure, we often need to solve large-scale linear systems," Chen says. "Liqiang's PLSQR work helps us solve those linear systems more efficiently."
The NWSC is the result of a partnership among NCAR; the University of Wyoming; the state of Wyoming; Cheyenne LEADS; the Wyoming Business Council; Cheyenne Light, Fuel and Power; and the University Corporation for Atmospheric Research. NCAR is sponsored by the National Science Foundation (NSF).
The NWSC will contain some of the world's most powerful supercomputers (1.5 petaflops, which is equal to 1.5 quadrillion computer operations per second) dedicated to improving scientific understanding of climate change, severe weather, air quality and other vital atmospheric science and geo-science topics. The center also will house a premier data storage (11 petabytes) and archival facility that holds irreplaceable historical climate records and other information.
Liqiang Wang, a UW associate professor of computer science, conducts research that includes optimizing use of the supercomputer in Cheyenne and improve the reliability of HPC (high-performance computing) programs. Wang will partner with UW colleague Po Chen on Chen's seismology research.