Supercomputing in Paradise

Paw Prints: Writings of the maddog

Imagine working in one of the most beautiful places in the world, high on a bluff in front of the magnificent Flatirons in Boulder, Colorado. Also imagine working in pure research dedicated to helping us understand the weather and climate that make up our environment, and finally imagine working with some of the most powerful supercomputers in the world. You might imagine that you are working for the National Center for Atmospheric Research (NCAR), and you would be right.

I was fortunate enough to speak in front of a group of researchers there, and before my talk they took me on a tour of their computing facilities and showed me a bit of what they are doing in research.

First of all, NCAR is not a government agency. It is a non-profit organization that receives funding from various government agencies to do pure research. It is primarily funded by the National Science Foundation and some other government agencies such as NOAA.

During World War II radar was developed, and the use of radar weather made vast advances in the predictability of weather. After World War II ended, a lot of these advances languished, and various universities studying weather and climate change slowed in their work. These universities came together to form a non-profit group called the University Corporation for Atmospheric Research (UCAR), which led to the founding and funding of NCAR.

Accurate weather forecasting, of course, has huge influences on our lives. Farmers, government, military and the average person all depend on having accurate weather forecasts, as well as being able to understand how the climate is changing and the various effects that elements such as the sun, the earth, pollution and historical trends have on our changing climate.

NCAR's mission is not only to do this research, but to disseminate their research as broadly as possible to whoever wants it. In addition, universities doing research in weather and climate can utilize their resources and data banks.

NCAR's scientists not only study the current climate, but through rock samples, ice bores, and other means determine what the climate has been in the past, and through computer models based on real-world measurements try to predict what the climate will be in the future. NCARS researchers also develop new tools, both hardware and software, to study the environment, then license them out to private companies to manufacture them for NCAR's use.

I enjoyed seeing the public area where NCAR had many exhibits showing what they have learned and how the various parts of the complex puzzle affect climate (and an explanation of why the chaotic environment make weather forecasting “difficult”), but I particularly enjoyed seeing the Computational and Information Systems Laboratory (CISL) where their supercomputers were.

They have three main supercomputers. One is called Bluefire, and is their 77-teraflop IBM Power6 system running AIX with an Infiniband Interconnect, one is call “Frost”, an 8,192 processor Blue Gene/L system running Linux 2.6, and the third main computing facility is “Lynx”, which is a Cray XT5m running Compute Node Linux which is based on SuSE Linux SLES 10. They also have clusters of computers for visualization, data storage and a high-performance data storage and archiving system that stores 12 Petabytes of information.

Even with all of this computing horsepower, it is not enough, and NCAR is building another supercomputer center, but this one will not be in Boulder. It was determined that it would be cheaper both from building costs and operating costs to have the new supercomputer center in Cheyenne, Wyoming. Once the building is finished and the supercomputer is in place, it is the goal of the center to be placed in the top 25 fastest computers in the world, twenty times more powerful than the current systems, with an energy usage of three to four megawatts, of which 10% will come from wind power.

Sometimes we concentrate so much on the mechanisms of creating Free Software, that we lose sight of some of the really great and important work that can happen because of it.

Davide Del Vento, a software engineer in the Consulting Services Group of the High-End Services Section (CISL), was my host and guide to the computing facilities as well as arranging my talk at the facility, and Scott Briggs helped with the travel arrangements. I wish to thank them and the NCAR management for making the trip possible.

Carpe Diem!

Comments