Organisers

James Annett, Pete Beckman, Bruce Boghosian, Shantenu Jha

The potential of distributed systems to engender new scientific results and methods is immense, but developing applications that can utilise rapidly evolving and complex infrastructure remains a challenge. Similarly, the next generation of new high performance computing (HPC) machines lead to many new challenges in software development and algorithm strategies. HPC machines are already exceeding the petaflop scale, opening up unprecedented scientific opportunities in computational modelling for global climate modelling, materials science, chemistry, and biomolecular science.  For example IBM's BlueGene/L has 130,000 processors distributed over 65,000 nodes. Programming strategies for such machines can be radically different from desktop workstations or even current HPC machines.

The aim of this workshop is to compile and characterise a range of computational science applications that have successfully exploited distributed and/or massively parallel infrastructure in challenging and novel ways to produce impactful and inspiring domain-specific results.  By showcasing successful applications spanning the scale of networking, computing and data resource requirements, we also hope to highlight the potential of distributed infrastructure and to assess future directions in high performance and distributed scientific applications.

This workshop will bring together leading scientists who are exploiting the boundaries of high performance and distributed computing. This workshop invites submissions from computational researchers with use cases that have successfully utilised distributed and/or high performance infrastructure. Whereas in the past it might have been possible for computational scientists to avoid parallel computing and obtain desired performance on their applications, with the advent of multi-core and related technologies, this is no longer an option.  For this reason, computational scientists who wish to maintain any kind of competitive advantage in the coming years are going to be forced to turn to parallel and distributed computing. Papers that describe the tools and techniques used to overcome these traditional barriers as well as a discussion of the novel results produced as a consequence will be of special interest to this workshop. Additionally, papers on tools and techniques that have been developed for grid-enabling applications are also welcome; contributed papers will be expected to provide an analysis of why distributed resources were required and how the use of distributed resources enabled results that would not have been possible otherwise.  As well as highlight scientific results the meeting will discuss parallelization strategies, code development strategies, and data handling and storage for the large data volumes from modern simulations and experiments such as LHC.

This workshop invites papers on topics that include but are not limited to:

• Case studies of new scientific work made possible by high performance and
  distributed computing
• Applications that exploit dedicated networks and optical lightpaths
• Applications utilising novel distributed algorithms
• Applications with challenging deployment and run-time requirements such as cross-grid information services
• Case studies where high-throughput, ensemble computing and/or integrating  computational resources from desktop to supercomputing have engendered new scientific insight
• Data-intensive applications
• Novel tools, techniques and infrastructure that assist the development, deployment and execution of distributed applications
• Methods frameworks, compilers, solvers, algorithms – to facilitate petascale applications
• Programming models for HPC/petascale applications.