A computational cluster and lab for
physics instruction and research
Brian E. Granger
Department of Physics
Department of Education
Center for Science, Technology and Society
March 4, 2004
Computers are increasingly a part of undergraduate education and faculty
research. This means that large amounts of University money have been spent, and
will continue to be spent, on computer hardware and software. Unfortunately, many if
not most computers on campus sit idle much of the time (overnight for instance). This
situation is especially tragic given the need of scientists for more and more
computational power to explore the today’s challenging scientific questions in areas
such as nanoscience and biotechnology, where large computer simulations can take
weeks or even months to complete on even the fastest computers. The principle author
of this proposal, Brian Granger, is a faculty member in the department of Physics at
Santa Clara University (SCU). His research involves computationally intensive
simulations of a number of systems from atomic and molecular physics as well as
With this background in mind, the goals of this project are to: 1) create a “mini-
supercomputer” at SCU needed to perform computational physics research in the group
of professor Brian Granger, 2) maximize the usage of computational resources already
existing on SCU’s campus, and 3) enhance or create computer labs in two academic
departments (Education and Physics).
Description of the project
The proposed computational resources will consist of two computer labs in the
departments of Education and Physics, as well as two additional computers in the
faculty office of professor Brian Granger. Using software written by Apple Computer
and others, all of these computers (about 16 total) will be linked together into a small
supercomputer or computational “grid.” The computers in this cluster will work together
to solve computational problems in physics that are too large to solve using a single
computer. The peak theoretical performance of this cluster will be about 125 GigaFlops
(1 Flops = 1 floating point operation per second). To put this number in perspective a
125 GigaFlops supercomputer would have been amongst the fastest 500
supercomputers at the end of 2001.
This dual usage of the computers for both student instruction and faculty
research provides a cost effective way of creating a powerful computational resource at
SCU and maximizing the usage of already existing computer resources. Because the
two computer labs in the departments of Physics and Education are at the center of the
proposed project, we now give a brief description of each of them.
Electronic Portfolio Laboratory (Bannan Hall 211). Currently, the department
of Education has a computer lab in Bannan 211 consisting of eleven G4 1.42 GHz, dual
processor Macintosh computers. This lab was established in 2003 by professor Pedro
Hernandez-Ramos with a $157,000 grant from the Fletcher Jones Foundation and two
grants (both during 2003) from the State of California for $8,500 each. During the day
this lab is used for courses in the department of Education, Libral Studies, Theater, and
others (courses vary from quarter to quarter). Currently, however, the lab sits idle
during the nights, weekends, and holidays.
Physics Computation Lab (Daly 308). The Department of Physics, under the
direction of Brian Granger, is currently creating a computer lab in Daly Science for use
in all upper division physics courses and certain general education physics courses
(The Physics of Dance). When completed, this lab will consist of approximately four
computers: three G5 2.0 GHz dual-processor Macintosh computers and one Linux-
based Dell computer. The funds for purchasing the computers and remodeling the
facilities for this lab are being provided by the Physics department and an SCU
University Research Grant. Two additional Macintosh computers located in the faculty
office of Brian Granger (Daly 312) will also be a part of the computational resources.
These computers were purchased with startup funds from the College of Arts and
Sciences at SCU.
The basic idea of this project is to tie all of these computers together with a high-
speed network and specialized clustering software to create a small supercomputer.
This supercomputer will then be available for physics research during periods when the
labs are not being used by students (nights, weekends, holidays). This approach
insures that the existing computational resources at SCU are used as much as possible.
Furthermore, the student computer labs in Physics and Education will also benefit from
the upgraded computers and network.
We have performed preliminary tests of the “supercomputer” to verify that our
idea is feasible. In these tests, we used Apple’s Xgrid product on the eleven computers
in the Education department lab. With this software, we were able to use all eleven
computers simultaneously to factor large numbers (basically a test program). These
tests also demonstrate the ability of the computers to seamlessly transition from large-
scale physics calculations to being used by students.
Our project will have significant educational impacts. More specifically, the
creation/enchancement of the two computer labs, and their linking up into a small
supercomputer will impact students in four departments: Physics, Education, Liberal
Studies and Theatre and Dance.
First, the lab in the physics department will enable computational methods to be
taught to all physics and engineering/physics majors. The curriculum in the physics
department is currently being modified to introduce students to computational methods
during their sophomore year. Then, these methods will be incorporated into all upper
division physics courses, providing the students with practical experience using state-of-
the-art computers to solve physics problems. This experience is critical for students
heading into industry and graduate school. The software requested in this proposal
(Mathematica and MatLab) is critical for creating these educational opportunities in the
Furthermore, professor Rich Barber in the physics department will be teaching a
“Physics of Dance” course. The students in this course will be primarily Theatre and
Dance majors. These students will use the advanced video processing capabilities in
the physics computer lab to analyze the physics of their own dancing. Without the
physics computer lab, another computer lab with advanced video processing
capabilities would need to be found or created from scratch.
Finally, students in the department of Education will benefit from the upgraded
capabilities of the computer lab in Bannan 211. With the increased RAM and high
speed network, students will be able to complete advanced projects in areas such as
video and photo editing, graphics creation and editing and stage design.
The primary reason for creating a small supercomputer at SCU is for the
computational/theoretical physics research of Brian Granger. This research focuses on
the many unanswered questions in molecular and atomic physics created by recent
experimental advances. Examples of this research include studies of cold antihydrogen
atoms, low-dimensional many body systems and the interactions of electrons with
nanoparticles. Simulating these systems on computers requires large amounts of
memory (multiple GB of RAM) and fast computers running for long periods of time.
The creation of the proposed computational grid will completely transform this
research program by providing a powerful computational resource capable of
performing large-scale simulations of novel physical phenomena. Without these
capabilities, such research is simply not feasible. Furthermore, having such a facility on
campus at SCU will provide students first hand experience in using such a computer
both in classroom settings and in faculty-advised research.
The budget requested from the Technology Steering Committee for this project totals
$38,384. This amount is a small portion of the total $245,000 budget for the project, as
other University and external sources have contributed large amounts of money towards
this project. A summary of this cost sharing is:
1. $174,000 in external grants for the establishment of the computer lab in Bannan
2. $17,000 in startup funds for Brian Granger’s research from the College of Arts
and Sciences, used to purchase the two computers in his office as well as an
array of scientific software.
3. Approximately $10,000 from the department of Physics for the purchase of two
computers for the Physics Computation Lab and the remodeling of the facilities
needed for this lab.
4. $5,612 from a University Research grant to purchase one computer for the
Physics Computation lab.
5. A flexible amount (up to a few $10k) from the Information Technology department
to upgrade the campus core network to support the Gb, high-speed network.
Thus the total funds contributed to this project from other sources is
We now provide a description of the budget requested of the TSC. To complete
this project, funds are needed for three purposes: memory upgrades, software and
Memory. Currently most the 11 of the computers in both labs have only 512 MB
of RAM. To be useful in large-scale physics simulations, these computers need to be
upgraded to at least 2GB of RAM. Without memory upgrades, the computers will have
a limited usefulness for physics research. Furthermore, students using the labs will also
benefit from the increased memory when performing computationally intensive projects.
Software. A number of software packages are needed to complete the project.
First, all of the computers need to be running the same, up-to-date version of the
Macintosh operating system (currently Mac OS 10.3.2). This both simplifies system
administration and allows highly optimized, custom physics simulations to run on all the
computers. To address this need we have requested money to upgrade all of the
computers in the cluster to the latest version of the operating system. Second, each
computer needs to run software that will enable all of the computers to cooperate as a
single supercomputer. We will be deploying Apple’s “Xgrid” (provided free of charge by
Apple) and Dauger Research’s “POOCH” for this purpose. We have budgeted funds to
purchase this software for each computer in the cluster. Third, software is needed for
the physics education lab (Mathematica and MatLab) which will be used in all upper
division physics courses to teach students to solve physics problems using
computational approaches. Without this software, students in the department of
Physics will not have the opportunity to learn this material.
Network infrastructure. When supercomputers are built using a number of
smaller computers, it is critical to have high-speed network connections between the
computers. This is required as the individual computers need to communicate rapidly
with each other during simulations. This communication enables the individual
computers to act as a single, effective supercomputer. By contrast, with a slow network
the computers often have to sit idle, waiting for information from other computers to
arrive. To address this need we propose to put all or most of the computers on a high-
speed network (preferably 1Gb). (The 11 computers in the Electronic Portfolio Lab are
already capable of connecting to 1GB networks.) An additional benefit of having a high-
speed network in the student computer labs in that system administrators will be able to
boot individual computers off the network, providing enhanced security and simplified
The details of this part of the budget are highly dependent on technical details of
the network upgrade, which have not been worked out yet. Discussions are currently
underway with Todd Schmitzer and others in IT to determine the best approach to
upgrading the network. Because of this, on the itemized budget, we quote a couple of
cost scenario’s for upgrading the network. At this point, we wish to emphasize that the
most costly scenario of having all the computers linked in a Gb network is strongly
desirable because of the effect it will have on the ability of the supercomputer to perform
certain types of calculations. While the less expensive options would still allow certain
types of physics calculations to be performed, they would dramatically limit the
capabilities of the supercomputer. Finally, this network upgrade provides a significant
enhancement of the campus network infrastructure. These enhancements are being
designed so they can be integrated with future network upgrades.