As a result of this award NSF will partially fund an international workshop on Clusters, Clouds and Grids for Scientific Computing. This workshop hosts 60 attendees and will take place in the fall of 2010 in North Carolina. The topic is relevant to the cyberinfrastructure for U.S. science and engineering researchers.
discussed the unprecedented transformation taking place in cluster, cloud and grid computing brought about by the emergence of multicore and hybrid microprocessor designs and by the escalation in the amount of data that leading edge scientific applications either generate or must assimilate and analyze. At this workshop, leading edge researchers in the field engaged in a sustained, creative dialogue that produced the kind of coordinated review and analysis necessary to translate such momentous developments into momentous benefits for the scientific community. Intellectual Merit of the Proposed Activities To achieve this goal, our workshop focused on both the commonalities that exist between clusters, clouds and grids and explored the ways they must interface and interoperate with one another to support today’s data intensive collaborations. The architectural similarities between these basic types of computational infrastructure, and the fact that high performance clusters typically make up the major compute nodes of grids and clouds, means that deployment, operational, and usage issues surrounding the latter form a superset of the issues surrounding clusters. Examples of topics explored include the following: overcoming scalability challenges (in several dimensions) that must be addressed to reach petascale performance and beyond; developing software that can scale both up and down the platform development chain, utilizing grid and cloud resources as necessary; agressive management of power consumption; achieving reliability and fault-tolerance for systems with thousands of processors; designing dramatically easier programming paradigms for highly parallel systems; creating tools (e.g., debuggers, performance monitors) that can support application development in highly virtualized environments, such as computing clouds, where the complexities of the substrate are hidden; and solving the numerous system management problems (e.g., system software, scheduling, security) that will be exacerbated by the new world of multicore architectures, resource virtualization and data intensive applications. Special attention was given to how these three major types of computational infrastructure can be integrated into a coherent whole that the community can easily use. Broader Impacts Resulting from the Proposed Activity In addition to providing a forum for dialogue between leading edge researchers in the field, we are producing two sets of journal articles based on talks given at the meeting. Each set will be published in separate special issues of well known journals in the fall of 2011. In addition to inviting a list of promising junior faculty to this year’s workshop, with help from Valerie Taylor, Texas A&M University, and Shirley Moore, University of Tennessee, we also worked to increase the participation of women and underrepresented groups at this meeting. Consequently, as compared to the previous meeting in this series (2008), we nearly doubled (from six to eleven) the number of women who attended.
