The Astronomy and Physics participation in the Millenium project has two main components. Twenty desktop workstations will be used for several purposes in Physics, with Fast Ethernet (FE) network availability to Astronomy. A twenty node NOW will be installed in Astronomy, with FE access to Physics.

Desktop Use and Progress

Eighteen of the desktops, running LINUX, all with 128MB memory, are being assembled as the "Physics and Astronomy Network of Intel Computers", PANIC. The network is described online (http://panic.millennium.berkeley.edu). The primary use is by first and second year graduate students who have yet to enter research groups. They are using the system for classwork, teaching and any research they might be undertaking. The first large-scale use of the machines will occur in Fall 1998. Nine of the desktops are up and running using standard ethernet in room 426 Birge, and are available to all comers; this step was completed in July 1998. The remaining nine will be installed in room 375 Birge in the next quarter. The FE networking for Physics is on order, and should be installed in the coming quarter.

Two of the desktops are in use by the AMANDA project, the construction of a very high energy neutrino observatory at the South Pole. One workstation is physically at the South Pole, and was used in a very successful test of an MBONE connection to the Pole. The connection was used to facilitate a live videoconference between schoolchildren in Mississippi and the winter-over personnel at the Pole; this was actually shown on PBS as part of their "Passport to Knowledge: Live from the Poles" special. More information on this may be found at http://quest.arc.nasa.gov/antarctica/

The other AMANDA Intel workstation is now an experimental server, used to do several things: 1) test codes before using them on the NOW, 2) constructing SQL databases to track the properties of all the optical modules in AMANDA, and 3) test secure ways of serving requests for data analysis and reconstruction to the AMANDA collaboration.

More information on AMANDA may be found at http://amanda.berkeley.edu/

Network of Workstations Parallel Computer

The first component of the Astronomy and Physics NOW, an Intel quad, was delivered in fall 1997 and installed on the network in room 544 Campbell, a temporary location. When the remaining components of the NOW are delivered, the whole assembly will be put together in room 47-6 Campbell, an air conditioned machine room in the Campbell basement. The existing machine, moon.berkeley.edu, has FE connection to the Astronomy subnet within floors 4-7 of Campbell Hall. CNS has completed the FE connection to room 47-6 in the basement, and is in the process of completing the upgrade of the Campbell Hall link to Evans Hall and to Physics, which also will go through room 47-6, work that should be done by the end of August. A 60GB RAID disk tower donated by IBM has been added to the 37GB disk space already in the machine, and a DLT tape drive has been added for backing up both the NOW and the PANIC cluster.

The Intel quad runs under SolarisX86, and has a collection of development tools, including Sun's C and Fortran compilers and development environment, the GNU compilers and development environment (both missing Fortran 90), MPI, PVM, Tex/Latex and CVS/SCCS/RCS. The rest of the desired software awaits Solaris x86 ports or purchases, and purchases await a more systematic organization of licensing of software for the Millenium project.

A number of science projects have been completed or are in progress on the existing 4 CPU machine. Peter Mardahl (EECS), working with Wurtele (Physics), has parallelized the fully relativistic, electromagnetic particle-in-cell plasma simulation code XOOPIC EM on the NOW with data on speed-ups obtained. Wurtele and Birdsall (EECS) plan on developing a moving-window PIC code, which has a window frame a few electromagnetic wavelengths long moving with v~c, for use in laser plasma interaction research. Arons (Astronomy and Physics) and Melatos (Astronomy Miller Fellow) are just starting on simulations of relativistic two stream instabilities and Landau masers of the ordinary mode in strongly magnetized electron-positron plasmas, in an investigation of the mechanism of pulsar radio emission.

AMANDA, the international high energy neutrino observatory project in which Price and Lowder in Physics are participating, has done a first-level filter of its entire data set from 1997 (about 0.5 terabyte). The NOW will be employed to do second-level filters and reconstruction on the filtered data, as well as searches for correlations between gamma-ray bursts and AMANDA events. This usage is in the planning and testing stage.

In conjunction with CS267, graduate students Fisher and Spitkovsky (Physics) designed and implemented a parallel Monte Carlo radiative transfer method in MPI, which can be run on both the local NOW and the CS NOW. A writeup describing the physical problem and the solution method can be found at

http://astron.berkeley.edu/~mookie/cs267/fp/267final/267final.html

Fisher (Physics), Klein (Astronomy) and McKee (Physics and Astronomy) have recently ported over various parallel Adaptive Mesh Refinement post-processing data tools developed by the NERSC CCSE group. While the basic AMR code with gravity in use by this group to study the hydrodynamics of star formation is still serial, these tools will help in visualizing and analyzing the results of the serial code (run on the C90 at NASA Ames NAS and the T90 at NPACI). In particular, for long-time series, both the amount of information stored and the amount of computation done in the post-processing stage can be quite significant, and parallelization on NOWs is an effective way of efficiently analyzing this information

The Cosmic Microwave Background group in Physics (Richards, Jaffe, Hanany) has yet to use the NOW for analysis of their data, but will have a large data set from the MAXIMA ballon experiment during the coming quarter, which they will use to test the very large matrix inversion algorithms required, a project for which the NOW is ideal (especially if the full twenty processors are available).