Presentation Abstracts
Main | Schedule | Details | Sponsors
Mellanox Technologies
Maximizing Productivity for Technical Computing
The increasing demand for computing power in scientific and
engineering applications has spurred the deployment of
high-performance computing clusters. With the growing size of clusters
and numbers of CPU cores per cluster server node, not only do the
traditional demands from the cluster interconnect increase
dramatically, but new demands are introduced. The cluster interconnect
needs to provide balanced throughput and latency to address the IO
requirements of each CPU core and to maintain high network utilization
for maximizing highest productivity. Mellanox Technologies, the leader
of industry-standard, high-performance InfiniBand and 10Gigabit
Ethernet solutions, provides a commodity, high-performance
interconnect fabric, which maximizes the technical cluster utilization
and efficiency. The usage of Mellanox interconnect solutions with
gridMathematica provides the most productive compute systems
for research, engineering, and analysis applications.
Speaker: Gilad Shainer
Microsoft and UnRisk Consortium
Dell
Financial Real-Time Risk Management as an Application of
Mathematica on HPC
Microsoft HPC provides a scalable platform for applications built on
gridMathematica. An example of one such application is UnRisk
FACTORY. After introducing Windows HPC, we will present the basic concepts
of the UnRisk FACTORY as a high-performance solution for financial risk
management. These are the main components of the UnRisk FACTORY:
- The UnRisk database holds termsheet information on
structured financial instruments, historical valuation results under
user-defined scenarios, and schedules for valuation.
- The UnRisk Service is a coarse-grain-parallelization of the
valuation tasks, based on gridMathematica.
- The UnRisk adapter is the interface to market data held
either in proprietary databases of the customer or provided by information
providers.
- A scalable number-crunching force of UnRisk engines
running under Microsoft Windows or Microsoft Compute Cluster carry out
the valuation itself.
Dell Computers Ltd., in conjunction with Microsoft and UnRisk, has
created preconfigured solution packages that link standards-based
servers together to act as a single, powerful compute engine. We will
also discuss how, by using off-the shelf, standardised components,
Dell customers can expand their clusters easily with new technologies
to meet increased demand while reducing the risk of a more proprietary
solution.
Speakers: Michael Newberry, Microsoft, and Andreas Binder, UnRisk
Consortium
Platform Computing
High-Performance Computing Made Simple
Distributed computing and cluster technologies have become part of
mainstream computing. This enables scientists, engineers, and industry
experts to rely on the ability to rapidly process complex datasets to
solve computation- and data-intensive problems. While the availability
of grid and cluster technologies has brought considerable capabilities to
users, the increased administration efforts and programming complexity
required by these technologies have significantly reduced their ease of
use. For most organizations, the challenge is to make the powerful
computational resources and capacity of grid and cluster computing
available to their users.
Platform Computing makes high-performance computing simple and affordable
with a family of products for cluster installation and parallel processing,
as well as workload, job, and task management for distributed computing
environments. This presentation will describe the implementation of
cluster technology and how Platform products integrate with
gridMathematica.
Speaker: Chris Dudding
Sun Microsystems
An Infrastructure for High-Performance Computation with
Mathematica 6
High-performance computing, once the domain of exotic CPU architectures
and arcane programming practices, is fast becoming a staple of commercial
computation. Commodity systems, tightly integrated middleware, and
significantly more capable application software facilitate the solution
of academic, scientific, and commercial problems that can generate new
business value. The changing nature of today's compute infrastructure will
be examined in light of the latest capabilities for parallel computation
found in Mathematica 6.
Speaker: James Coomer
