This talk will introduce the Cray J-90 systems. There are six such systems, five of which are dedicated to batch use. The system architecture, software, and usage modes will be described.
NQE and NQS: effective use of the NERSC Batch system at NERSC. A good explication of the queuing systems, the queues, and scripting issues useful to the batch user. How to use the T3E and J-90s, via batch jobs, from development, through debugging, and into production.
This talk will describe the file systems relevant to NERSC computer systems. These include home directories, scratch space, mass storage, and networked file systems. Advice will be given for efficient use and resource conservation, with examples to illustrate useful techniques. Reference details of commands and utilities will also be provided.
One of the most serious threats to current day computer security is "sniffing" of clear text passwords from shared communication lines. The Secure Shell (SSH) software can protect against this threat by providing encrypted connections (e.g. "Telnet" connections). Incorporated into SSH are facilities such as IP "tunneling" (e.g. for FTP or X-Windows sessions) and certificated client authentication. This talk describes SSH operation and use in the context of communication with NERSC, including examples from the most popular desktop platforms (PC, Mac, Unix).
Scientific Visualization Using NCAR Graphics
[Presentation not available]. NCAR Graphics is a collection of graphics libraries that support display of scientific data. It has utilities for contouring, mapping, drawing field flows, surfaces, histograms, X/Y plots, labeling, and more. It includes a `Pictorial Guide to Examples' that illustrates a wide range of graphical presentations developed and employed at NCAR. The current version of the package also includes a math library.
Parallel I/O and parallel file management on the T3E. How to do single-file and multiple-file I/O, using the facilities Cray provides; will include discussion of MPI I/O, which is now available on the T3E. Performance pragmatics, as well as practicality issues (# files, # open files, # inodes, data quantity, I/O bandwidth, mass storage, etc.), and real performance numbers if available.
This session will be a brief overview of what defines a scientific visualization, examples of successful visualizations and a review of some of the resources available at NERSC for scientific visualization. Then we will describe the basic methods of creating short digital movies of scientific visualizations, potential pitfalls to avoid, and tools available at NERSC to create movies.
The MPI Forum reconvened during 1995-97 to consider extensions to the MPI message-passing interface standard that had been deliberately put aside during the successful MPI-1 definition process in 1992-1994. Extensions that were defined include standards for parallel I/O, remote memory operations (put/get), and dynamic process management. Now nearly all systems support the MPI-2 standard for parallel I/O, and various vendors and public implementors are in the process of delivering the rest of the MPI-2 specification. This talk will describe the salient features of MPI-2 and how they can be effectively used by applications. It will also summarize the status of MPI-2 implementations by vendors. [Argonne National Laboratory]
This talk will describe methods for achieving high performance on the J-90 systems. These are shared memory vector multiprocessor systems, and it is easier to achieve a high percentage of peak performance from them than other architectures. We will describe vectorization and parallelism, and how to achieve them with the combination of careful code design and compiler directives.
Fortran-77 and Fortran-90: Porting Between PVP and MPP
[Presentation not available]. This lecture will be an overview of where Fortran 90 fits into the current selection of modern computer programming languages, and the model it encourages. We will show Fortran90 examples that of possible replacements for some common Fortran77 idioms. we will also introduce designing and coding for parallelism. Intended for those new to parallel computing, it will be a presentation of concepts and broad general principles that characterize and differentiate parallel processing on shared memory and massively parallel computers.
A Briefing on the NERSC III System
[Presentation not available]. A technical description of the newly announced NERSC III system: what it will consist of, the phases of its delivery, and a discussion of how it will be used and managed. Both hardware and software will be discussed.
The ACTS talk was not given at this event; a briefing on the NERSC III system was given, instead. The links below are to the ACTS presentation material that would have accompanied this talk.
Coverage of the specialized libraries and solvers in the ACTS Advanced Computational Testing and Simulation) toolkit that can help implement and manage parallelism. A survey of the state of the tools, what they do well and poorly, and possibly some example uses.
The development of application codes to simulate physical phenomena is complex, time-consuming, and requires specialized expertise. Writing parallel codes from scratch is often beyond the resources of any individual or small team. The PETSc toolkit provides a rich set of data structures and algorithms to ease the development of scalable PDE-based simulation codes. PETSc provides a variety of parallel linear and nonlinear algebraic solvers as well as much of the software infrastructure to manage mesh data structures, without requiring one to program explicitly in MPI. This talk will introduce the organization and use of PETSc for the parallel solution of PDEs. [Argonne National Laboratory]
Emerging high-performance networks promise to enable a wide range of emerging application concepts such as remote computing, distributed supercomputing, teleimmersion, smart instruments, and data mining. However, the development and use of such applications are, in practice very difficult and time consuming, because of the need to deal with complex and highly heterogeneous systems. The Globus grid programming toolkit is designed to help application developers and tool builders overcome these obstacles to the construction of "grid-enabled" scientific and engineering applications. It does so by providing a set of standard services for authentication, resource location, resource allocation, configuration, communication, file access, fault detection, and executable management. These services can be incorporated into applications and/or programming tools in a mix-and-match fashion to provide access to needed capabilities.
In this talk, I introduce the capabilities of the Globus toolkit and describe how Globus services can be applied in specific applications. [Argonne National Laboratory]
The design of optimized fusion plasma confinement devices is a topic of significant current interest in magnetic fusion energy research. This takes its most general form in stellarator configurations which have fully 3-dimensional magnetic flux surface geometries and coil shapes. The numerical design of such devices relies heavily on high-performance computing and scientific visualization. In addition, the analysis of plasma confinement in these devices is most naturally carried out using parallel Monte Carlo particle simulation techniques on massively parallel computers such as the T3E. Our techniques and experiences using parallelization on the T3E and scientific visualization of the results on local workstations will be described. [Oak Ridge National Laboratory]