NCAR plays a critical national role in the development and provision of effective, end-to-end cyberinfrastructure (CI) for the atmospheric and Earth sciences community. NCAR's robust, innovative, and accessible supercomputing and data services have made customized high-end resources available to thousands of scientists and students, supported the creation and ongoing improvement of world-class atmospheric models, and enabled numerous cutting-edge simulations.
This track record of leadership, service, and success will continue. We will marshal the necessary facilities, equipment, software, and numerical and intellectual capabilities to meet the ever-expanding scientific goals of the geosciences community. In particular, we anticipate a 20-fold increase in the atmospheric community's demand for cyber-resources by 2012-requiring a capability comparable to the NSF "Track-2" or "petascale" centers. To be effective, this Track-2 -scale supercomputer must also be connected to a balanced and customized set of data systems, including petabyte-scale high-performance file systems shared with advanced data analysis and visualization resources, all backed up by a hundred-petabyte mass storage system. The entire facility must be connected, via high-performance networking, to the nation's other leading CI facilities, particularly those centers operated by NSF and the Department of Energy. The power, space, and cooling requirements of this facility will certainly exceed the capabilities of the NCAR Mesa Laboratory. Constructing and beginning operation by 2012 of a new supercomputing facility capable of housing Track-2-scale systems is, therefore, NCAR's top priority. Over the next three to five years, NCAR will
Looking further ahead, higher wide-area-network bandwidth, more powerful computers, and specialized software are creating a grid that promises to provide simplified access to distributed high-performance computing resources. A grid strategy offers several potential benefits. By working together, geographically dispersed partners can scale up processing cycles to meet increasingly large scientific challenges. In addition, complementary expertise and resources can be combined across centers to tackle the complex and interdisciplinary challenges in computational science that are difficult or impossible for one center to address in isolation. These challenges include many computational problems in the geosciences, such as climate and earthquake modeling. Furthermore, the grid is part of a cluster of technologies forming a trend in high-end computing toward "cloud computing": system virtualization, center automation, service-oriented architectures, and distributed workflows. Keeping up with and even leading the evolution of these ideas is essential to the long-term health of the laboratory.