Ben Mar

CS 267 Assignment 0

1/26/06

About Me

I am a first year graduate student in Electrical Engineering at University of California, Berkeley. I received my BS in Electrical and Computer Engineering from Worcester Polytechnic Institute. My interests are computer architecture and reconfigurable systems. By taking CS 267 I hope to learn how to write parallel programs and gain some experience working with the powerful computers.

Space Weather Simulation using BATSRUS

The Sun–Earth system is an extremely complex natural system involving many different interacting elements. The goal of this project (Clauer, et al, 2000) is to develop a framework that can simulate and predict the properties of space weather such as solar wind, magnetosphere, ionosphere, and thermosphere. The problem is representing the large range of scales for critical phenomena in this system, which requires advanced numerical technology.

The advanced numerical code used was BATSRUS-Block Adaptive-Tree Solar-wind Roe-type Upwind Scheme and has been used successfully to simulate the environment from the solar surface to the top of the Earth’s upper atmosphere. BATSRUS was used as it could be easily implemented with parallel machines. The basic unit of data was a three-dimensional block of grid cells and the data in the block were stored in i,j,k-indexed arrays.

In the grid above, each level of refinement introduces cells that are smaller by a factor two in each dimension from those one level higher in the grid. Calculations can have more than 20 levels of refinement. The BATSRUS code was designed to optimize serial performance on single processors of the target machine while the overall adaptive block scheme also reduces communication to provide high parallel performance. The adaptive-block data structures that the code was built on allowed easy, natural partitioning of the data, and greatly facilitated load-balancing. Balance and efficiency were achieved by partitioning the blocks among the available processors.

The code was written in Fortran 90 with message-passing via MPI library. By using MPI they were able to test in a wide range of machines, from integrated shared-memory systems to networks of workstations.

The code looks to perform well with both Cray machines while differing slightly with the SGI. The biggest difference between actual performance and ideal performance was seen in the Beowulf.

The computers are not in the top 500 list today however when the study came out (in 2000) the Cray T3E-1200 ranked 10th.

The magnetic streamlines of the simulation produce a close model to the recorded observations. The two patterns are topologically identical. They both display two saddles and two nodes. The stable node is in the upper half plane which shows the magnetic streamlines that converge on the node and the unstable node is in the lower half plane which shows the magnetic streamlines that leaves the node. The location and orientation of the critical points are quite similar.

References

C. Robert Clauer, Tamas I. Gombosi, Darren L. De Zeeuw, et al. “High Performance Computer Methods Applied to Predictive Space Weather Simulations.” IEEE Transactions on Plasma Science, Vol. 28, No. 6, December 2000

Darren L. De Zeeuw, Tamas I. Gombosi, Clinto P. T. Groth, Kenneth G. Powell, and Quentin F. Stout. “An Adaptive MHD Method for Global Space Weather Simulations.” IEEE Transactions on Plasma Science, Vol. 28, No. 6, December 2000

C. P. T. Groth, D. L. De Zeeuw, K. G. Powell and T. I. Gombosi. “A Parallel Adaptive 3D MHD Scheme For Modeling Coronal and Solar Wind Plasma Flows.” Space Science Reviews, Volume 87, Numbers 1-2, March 1999, p193 - 198