Computing at the Pitzer Center

The Pitzer Center for Theoretical Chemistry contains state-of-the-art computer resources that serve the needs of the 7 theory groups. These are typically Linux or Unix clusters that consist of between 30 and 70 processors. A survey of the groups in summer 2004 showed that in aggregate, we have well over 500 processors in total!

The speed of most of these processors for performing basic floating point arithmetic is between 1 and 8 GFlops (each GFlop being a billion floating point operations). Thus we have collective capability of over a trillion adds and multiplies per second. Enough to do some very large-scale simulations!

The architecture of most of the clusters consists of racks of pizza-box size compute nodes, each containing typically 2 processors. The theory groups configure their nodes differently according to their different scientific computing needs. Thus memory and disk per processor range from relatively modest amounts up to 16 GB memory per node, and 1 TB disk per node!

Some clusters are also put together from tiny"blades" which permit higher density than the standard "1u" form factor pizza boxes. Linux is the predominant operating system, with processors ranging from Intel Xeons to AMD Athlons and Opterons. In addition, we also have clusters configured from Apple's G5 XServe nodes, and an IBM SP2 computer running AIX.

Within the Pitzer Center itself, there is both high-speed wired networking and additionally wireless networking. Theory groups provide desktop computers to access the clusters, as well as digital library content, graphics etc. Additionally most of the theory groups take advantage of the Department of Chemistry
Graphics Center, which provides standard software packages, large format printers, and additional computer resources.

Most of the theory groups also have access to supercomputer centers, including the NSF centers at San Diego and Illinois. The main supercomputer use is undoubtedly at the National Energy Research Scientific Computing Center (NERSC), which is based at Lawrence-Berkeley National Laboratory, adjacent to U.C.Berkeley. NERSC is the flagship scientific computing facility for the Office of Science in the U.S. Department of Energy and one of the largest facilities in the world that provides computational resources and expertise for basic scientific research. Their main machine is the NERSC IBM SP RS/6000 named Seaborg: a distributed memory computer with 6,080 processors capable of running massively parallel scientific applications. Each processor has a peak performance of 1.5 GFlops.