ACADEMIA
Q&A: Stanford and Sun’s Computational Earth and Environmental Science Facility
Center Aims to Drive Energy Savings, Develop Efficient Technologies and Distill Business Benefits -- Stanford University President John Hennessy and Sun Microsystems Chairman Scott McNealy opened the doors to the Stanford Computational Earth and Environmental Science (CEES) research facility today. The facility will be used to expand the present capacity for interdisciplinary Earth science research, and is expected to allow deeper analysis, simulation and prediction around complex Earth processes and systems -- which could lead to advances in earthquake detection, oil exploration and the effects of oceanic and climate changes. Supercomputing Online interviewed Joerg Schwarz, Director of Life Sciences and Healthcare at Sun. 
SC Online: Please provide a technical break-down of the new supercomputing system the CEES research facility will be using. Schwarz: The system consists of a Sun SF 6900 with 24 dual core UltraSPARC IV CPU and 192GB of main memory, which can be addressed as a single system image. This part of the compute environment is targeted at applications that benefit from very large memory, for example because entire data sets can be stored in memory, instead of swapping data between disks, cache and memory. There is also a cluster of 64 Sun V20 dual cpu, dual-core AMD Opteron Servers, with 4GB of memory each, plus four Sun V40 Servers with 4 dual core AMD Opteron cpus and 32 GB of memory each. The versatility of the environment allows to allocate the ideal platform for throughput and massive parallel applications, over scalable, yet memory intensive applications to the aforementioned very large memory applications. All systems can access a joint file system. The AMD Opteron Servers are certified for Windows, Linux and Solaris. The goal is to run the entire Center with it's heterogeneous platforms with the Solaris operating system and use the same tools for programming, provisioning and management. At the same time, Linux can be provisioned for legacy application for which source code re-compilation is not an option. SC Online: It seems like the environment to be installed will be somewhat mixed...Grid of x64 as well as SPARC-based Sun Fire servers. Is a mixed computing environment beneficial for this type of research? Schwarz: Yes, it is essential. The different architectures have different focus areas, just like different applications have different requirements. All too often, much energy is wasted by mapping applications to platforms that are not an ideal match. In this environment, the coupling of different applications running on the platforms that are ideal for them promises advances in many areas of scientific computing. SC Online: What do you consider to be one or two of the most significant advances in building a large-scale computational Grid for complex research? Schwarz: Single systems are insufficient to handle the demands for computational and data-management resources required by complex numerical simulations. Grid environments allow combining heterogeneous resources, for example compute clusters of homogeneous nodes and large memory systems, and utilizing the specific architectural features to the advantage of the entire, complex system. While many clusters and grids merely run older application software, the full power of Grids will be unleashed with a new style of coupled applications that dynamically detect and select resources based on fitness relative to the platform characteristics. Sun builds these sophisticated environments and middleware, while Stanford's CEES is a driving force in writing applications that utilize those. Together we can learn from each other and drive the common agenda forward, which will lead to new Grid utility environments and new types of grid enabled and optimized applications synchronously. Supercomputing Online wishes to thank Joerg Schwarz for his time and insights. Bridging the needs of large corporations with expertise from academia, the center also aims to tackle technology issues that plague the energy industry. For instance, researchers will study the detrimental effects of using applications with inefficient computing systems and explore business technology pain points related to data center consolidation and grid computing. "Our mission is to enhance the capacity for large-scale computational research for Earth and environmental science," says CEES director Jerry M. Harris, professor of geophysics in the Stanford School of Earth Sciences. "A driving force for this is the fact that, here at Stanford, we have some of the world's best scientists, and across the street in Silicon Valley are some of the world's best computer designers and builders." "Working with the energy sector for over 20 years, Sun is helping bridge the gap between academia and industry and by supporting the CEES. This provides a unique and innovative architecture for functional analysis of very complex applications used in Earth and environmental science," says Kim Jones, vice president of global education and research at Sun. "As part of Sun's Eco-Responsibility initiative, the company remains committed to innovating products to provide the energy industry with the most advanced computational resources to power the next generation of energy solutions for the planet, while building further efficiencies into existing processes." The CEES is a one of a kind facility built by Stanford, government agencies and private companies. Sun Microsystems contributed hardware and software, including its x64 (x86, 64-bit) and SPARC-based Sun Fire servers, and the Solaris Operating System (OS), and program support to aid this research facility. Sun is a founding member of what the university hopes will be a larger consortium. Other board members include representatives from Chevron and BP among others. For more information on CEES, please visit its Web site.
SC Online: Please provide a technical break-down of the new supercomputing system the CEES research facility will be using. Schwarz: The system consists of a Sun SF 6900 with 24 dual core UltraSPARC IV CPU and 192GB of main memory, which can be addressed as a single system image. This part of the compute environment is targeted at applications that benefit from very large memory, for example because entire data sets can be stored in memory, instead of swapping data between disks, cache and memory. There is also a cluster of 64 Sun V20 dual cpu, dual-core AMD Opteron Servers, with 4GB of memory each, plus four Sun V40 Servers with 4 dual core AMD Opteron cpus and 32 GB of memory each. The versatility of the environment allows to allocate the ideal platform for throughput and massive parallel applications, over scalable, yet memory intensive applications to the aforementioned very large memory applications. All systems can access a joint file system. The AMD Opteron Servers are certified for Windows, Linux and Solaris. The goal is to run the entire Center with it's heterogeneous platforms with the Solaris operating system and use the same tools for programming, provisioning and management. At the same time, Linux can be provisioned for legacy application for which source code re-compilation is not an option. SC Online: It seems like the environment to be installed will be somewhat mixed...Grid of x64 as well as SPARC-based Sun Fire servers. Is a mixed computing environment beneficial for this type of research? Schwarz: Yes, it is essential. The different architectures have different focus areas, just like different applications have different requirements. All too often, much energy is wasted by mapping applications to platforms that are not an ideal match. In this environment, the coupling of different applications running on the platforms that are ideal for them promises advances in many areas of scientific computing. SC Online: What do you consider to be one or two of the most significant advances in building a large-scale computational Grid for complex research? Schwarz: Single systems are insufficient to handle the demands for computational and data-management resources required by complex numerical simulations. Grid environments allow combining heterogeneous resources, for example compute clusters of homogeneous nodes and large memory systems, and utilizing the specific architectural features to the advantage of the entire, complex system. While many clusters and grids merely run older application software, the full power of Grids will be unleashed with a new style of coupled applications that dynamically detect and select resources based on fitness relative to the platform characteristics. Sun builds these sophisticated environments and middleware, while Stanford's CEES is a driving force in writing applications that utilize those. Together we can learn from each other and drive the common agenda forward, which will lead to new Grid utility environments and new types of grid enabled and optimized applications synchronously. Supercomputing Online wishes to thank Joerg Schwarz for his time and insights. Bridging the needs of large corporations with expertise from academia, the center also aims to tackle technology issues that plague the energy industry. For instance, researchers will study the detrimental effects of using applications with inefficient computing systems and explore business technology pain points related to data center consolidation and grid computing. "Our mission is to enhance the capacity for large-scale computational research for Earth and environmental science," says CEES director Jerry M. Harris, professor of geophysics in the Stanford School of Earth Sciences. "A driving force for this is the fact that, here at Stanford, we have some of the world's best scientists, and across the street in Silicon Valley are some of the world's best computer designers and builders." "Working with the energy sector for over 20 years, Sun is helping bridge the gap between academia and industry and by supporting the CEES. This provides a unique and innovative architecture for functional analysis of very complex applications used in Earth and environmental science," says Kim Jones, vice president of global education and research at Sun. "As part of Sun's Eco-Responsibility initiative, the company remains committed to innovating products to provide the energy industry with the most advanced computational resources to power the next generation of energy solutions for the planet, while building further efficiencies into existing processes." The CEES is a one of a kind facility built by Stanford, government agencies and private companies. Sun Microsystems contributed hardware and software, including its x64 (x86, 64-bit) and SPARC-based Sun Fire servers, and the Solaris Operating System (OS), and program support to aid this research facility. Sun is a founding member of what the university hopes will be a larger consortium. Other board members include representatives from Chevron and BP among others. For more information on CEES, please visit its Web site. 
How to resolve AdBlock issue?