Gordon

On November 4th, the San Diego Supercomputing Center (SDSC) announced that they have been awarded $20 million by the National Science Foundation to develop a new supercomputer aimed at "solving critical science and societal problems now overwhelmed by the avalanche of data generated by the digital devices of our era." This new computer is known as Gordon.

The SDSC has been part of the National Science Foundation's plan for cyberinfrastructure in the sciences for a number of years, as we read earlier in the semester. The development of Gordon is part of the NSF's continuing effort to keep up with the computing needs of the sciences. According to Jose L. Munoz, the deputy director and senior science advisor for the NSF's Office of Cyberinfrastructure, "'Gordon will do for data-driven science what tera-/peta-scale systems have done for the simulation and modeling communities, and provides a new tool to conduct transformative research.'"

Gordon, which is scheduled to be installed by Appro International, Inc. in 2011, will employ flash memory to speed solutions to data-intensive problems, such the analysis of individual genomes, much faster than spinning disk technology. Gordon is the follow up to the Dash system, another SDSC project, which was the first computer to use flash devices. Gordon will feature 245 teraflops of total compute power, 64 terabytes of DRAM, 256 terabytes of flash memory, and four petabytes of disk storage.

Another feature of Gordon will be 32 so-called "supernodes," each consisting of 32 compute nodes capable of 240 gigaflops/node and 64 gigabytes of DRAM. Linked together by virtual shared memory, these "supernodes" each have to potential of 7.7 teraflops of compute power and 10 terabytes of memory. The "supernodes" will be linked together with an InfiniBand network, which is capable of 16 gigabits per second of bidirectional bandwidth, which, apparently, is eight times faster than some of the most recently developed supercomputers. Gordon will be made available to researchers via an open-access national grid.

The point of Gordon is to make possible the complex applications of data-driven science. According to the press release, Gordon will have potential benefits in both the academic and industrial settings. Gordon should also be useful for predictive science, which seeks to make models of real-life phenomena. Because of the large-scale memory on a single node and the consequent increase in computation speeds, Gordon should allow for the creation of models that more accurately mimic these phenomena. With such capabilities, Gordon will be the next step in the development of data-driven science.