VISUALIZATION

The High Performance Architectural Challenge: Mass Market versus Proprietary Components? Guest Editors: Mark Baker (University of Portsmouth) and Daniel S. Katz (Jet Propulsion Laboratory) A special issue of Concurrency and Computation: Practice and Experience (CCPE) is being planned for the winter of 2003. Papers submitted and accepted for this issue will be published by John Wiley & Sons Ltd. in the CCPE Journal and in addition will be made available electronically via the Web. Background: The past two decades has seen a wide ranging and sometimes intense debate between parishioners of high performance and parallel computing about which system architecture will deliver the best performance for the science and engineering community applications. The recent trend in computer architectures has been a move away from dedicated and proprietary systems towards commodity off-the-shelf (COTS) systems. A leading example of this is the US Accelerated Strategic Computing Initiative (ASCI) whose objective is to meet the science and simulation requirements of the Stockpile Stewardship Program. For this purpose the US Dept. of Energy has funded large commodity based systems at its major laboratories (from IBM, SGI/Cray, and Intel). In contrast, Japan has funded the Earth Simulator. This project represents an attempt to build a supercomputing platform, which combines proprietary NEC SX vector processors with a fast hierarchical interconnect, that has the ability to handle the challenges of global fine-scale modelling of the Earth's climate and seismic activities. According to the TOP500 list the Earth Simulator has a peak performance of nearing 40 Tflop/s, compared to the largest US ASCI White system at Lawrence Livermore of just over 7 Tflop/s, a five fold performance advantage. Not only is the peak performance of the Earth Simulator high, three entrants to the Gordon Bell prize at SC2002 in Baltimore showed real application performance of between an impressive 12 and 26 Tflop/s. The highest performing of these codes used MPI, but it is interesting to note that one code used a variant of HPF (JHPF) and achieved very high performance. The emergence of the Earth Simulator has refueled the debate about not only high performance and parallel architectures, but also the most effective programming paradigm for these system. Call For Papers This is a general call for papers about the performance of the Earth Simulator and similar high performance parallel platforms. Particular areas of interest are those related to the implications of the system architecture and performance on such platforms. In addition, papers related to the use and implications of particular programming paradigm for applications in OpenMP, HPF or MPI, compared to other similar platforms will be of interest. Topics of interest for this special issue include but are not limited to: … Aspects of hardware and software performance, including I/O and visualization, … Comparative studies looking at the Simulator against other HPC platforms, … Features of HPF and MPI-2 that can enable very high performance and scalability, … The implication of using HPF or MPI-2 for programming applications, … System optimizations to aim application performance, … Methods of programming large coupled applications for particular hardware, such as the Earth Simulator Timescales for Submission There is a deadline of 1st February 2003 for submitted papers. Publication is currently scheduled for the fourth quarter of 2003. Activity Deadline Call For Papers 1st November 2002 Paper Submission Deadline 1st February 2003 Paper Approval 15th March 2003 Final Papers Approved 15th July 2003 Publication Late Q4 2003 Further details about this special issue can be found at: http://www.dsg.port.ac.uk/Journals/CCPE/Special/Earth-Simulator