A Personal Supercomputer for Climate Research

We describe and analyze the performance of a cluster of personal computers dedicated to coupled climate simulations. This climate modeling system performs comparably to state-of-the-art supercomputers and yet is a ordable by individual research groups, thus enabling more spontaneous application of high-end numerical models to climate science. The cluster's novelty centers around the Arctic Switch Fabric and the StarT-X network interface, a system-area interconnect substrate developed at MIT. A signi cant fraction of the interconnect's hardware performance is made available to our climate model through an application-speci c communication library. In addition to reporting the overall application performance of our cluster, we develop an analytical performance model of our application. Based on this model, we de ne a metric, Potential Floating-Pointing Performance, which we use to quantify the role of high-speed interconnects in determining application performance. Our results show that a highperformance interconnect, in conjunction with a light-weight application-speci c library, provides e cient support for our ne-grain parallel application on an otherwise general-purpose commodity system.

[1]  Charles L. Seitz,et al.  Myrinet: A Gigabit-per-Second Local Area Network , 1995, IEEE Micro.

[2]  G. Andrew Boughton.  Arctic Routing Chip , 1994, PCRCW.

[3]  Arvind,et al.  T: A Multithreaded Massively Parallel Architecture , 1992, [1992] Proceedings the 19th Annual International Symposium on Computer Architecture.

[4]  James C. Hoe,et al.  MPI-StarT: Delivering Network Performance to Numerical Applications , 1998, Proceedings of the IEEE/ACM SC98 Conference.

[5]  Scott Pakin,et al.  High Performance Virtual Machines (HPVM'S): Clusters with Supercomputing API's and Performance , 1997, PPSC.

[6]  J. Holton.  An introduction to dynamic meteorology , 2004 .

[7]  Richard P. Martin,et al.  LogP Performance Assessment of Fast Network Interfaces , 1995 .

[8]  L. Perelman,et al.  A finite-volume, incompressible Navier Stokes model for studies of the ocean on parallel computers , 1997 .

[9]  David E. Culler,et al.  An Implementation and Analysis of the Virtual Interface Architecture , 1998, Proceedings of the IEEE/ACM SC98 Conference.

[10]  Thomas L. Sterling,et al.  Scaling of Beowulf-class Distributed Systems , 1998, Proceedings of the IEEE/ACM SC98 Conference.

[11]  Andrew Shaw,et al.  A Comparison of Implicitly Parallel Multithreaded and Data-Parallel Implementations of an Ocean Model , 1998, J. Parallel Distributed Comput..

[12]  Chris Hill,et al.  Efficient ocean modeling using non-hydrostatic algorithms , 1998 .

[13]  R. Giering,et al.  Construction of the adjoint MIT ocean general circulation model and application to Atlantic heat transport sensitivity , 1999 .

[14]  A. Adcroft,et al.  Representation of Topography by Shaved Cells in a Height Coordinate Ocean Model , 1997 .

[15]  Thomas L. Sterling,et al.  BEOWULF: A Parallel Workstation for Scientific Computation , 1995, ICPP.

[16]  L. Perelman,et al.  Hydrostatic, quasi‐hydrostatic, and nonhydrostatic ocean modeling , 1997 .

[17]  John Marshall,et al.  Application of a parallel Navier-Stokes model to ocean circulation , 1996 .

[18]  Larry Rudolph,et al.  StarT-Voyager: A Flexible Platform for Exploring Scalable SMP Issues , 1998, Proceedings of the IEEE/ACM SC98 Conference.

[19]  James C. Hoe.  StarT-X - A One-Man-Year Exercise in Network Interface Engineering , 1998 .

[20]  Andrea C. Arpaci-Dusseau,et al.  Parallel computing on the berkeley now , 1997 .

[21]  G. Andrew Boughton.  Arctic Switch Fabric , 1997, PCRCW.