ISDP 2012 - International WORKSHOP on In-Situ Data Processing technologies

The last decade has seen substantial improvements in the capabilities of leadership computing systems, with the Top500 machines seeing a many-fold increase in computing power from a few TFlops to more than a PFlop of peak performance provided by the current generation of High End Computing (HEC) systems. This growth has allowed scientists to explore fundamental research questions and obtain dramatic new insights in science and engineering through very large scale and long-running simulations. However, these simulations are also generating an unprecedented amount of data. This presents significant challenges because the enormous increases in compute power have not been matched by corresponding increases in I/O performance.
Addressing I/O performance is a serious challenge for leadership class machines, in part because conventional methods for addressing I/O system bottlenecks, such as for example, increasing the I/O back-end capability by adding more disks with higher speeds, are unlikely to continue to scale to acceptable levels due to constraints imposed by cost and power. Coupled with the increasing data volumes and data rates associated with large scale simulations, this imbalance has resulted in a new bottleneck that is rapidly impacting the quality and productivity of the scientific research that these systems were designed to advance.
This workshop focuses on innovative approaches for scalability in the end-to-end processing and management of scientific data, with a particular interest in approaches that enable a holistic, rather than piecewise, approach. The workshop will bring together researchers from academic, industrial, and laboratory environments to discuss solutions for end-to-end data management for the next generation of scientific computing environments.