ARMONK, NY - 09 Mar 2006: IBM today announced that it has scored a fundamental performance breakthrough in the way that massive computer networks access and share information.
Code-named "Project Fastball," the historic results were achieved on the ASC Purple supercomputer -- the third most powerful supercomputer in the world.(1) ASC Purple, located at Lawrence Livermore National Laboratory, is a National Nuclear Security Administration computing system designed by IBM.
The breakthrough is expected to drive the development of data-intensive applications in areas including customized medicine, online gaming, entertainment, and homeland security, as well as in traditional high performance computing applications.
IBM and LLNL demonstrated over 102 gigabytes per second of sustained read and write performance to a single file using specialized software that manages the transfer of information between thousands of processors and thousands of disk storage devices. The world record performance was achieved using 416 individual storage controllers combined with 104 Power-based eServer p575 nodes.
The resulting file system was a remarkable 1.6 petabytes in size -- one of the largest high performance file systems currently deployed in the world. Demonstrating the scalability of the system, the performance was maintained as over 1,000 clients drove workloads to the file system.
"Advances at the upper reaches of high performance computing such as these often find powerful applications in the broad spectrum of government, industrial and commercial arenas," said David Turek, vice president of Deep Computing at IBM. "IBM's close working relationship with Lawrence Livermore National Laboratory has resulted in another significant achievement with the potential to improve the lives of people everywhere."
As the performance of the world's top supercomputers and commercial computing networks grows at an unprecedented rate, the need to manage the massive stores of data generated by those systems also increases.
Advanced Simulation and Computing (ASC) is a cornerstone of the National Nuclear Security Administration's (NNSA) program to ensure the safety, security and reliability of the nation's nuclear deterrent without nuclear testing. ASC unites the computational resources and expertise of Lawrence Livermore, Los Alamos and Sandia national labs.
For innovators in the field of supercomputing, the result redefines commonly held expectations for the boundaries of file system scalability in high performance computing deployments -- a critical area of importance as the world's top supercomputing systems move closer to functional performance levels of a single petaflop and more.
"The performance achieved on the ASC Purple system demonstrates the impressive scalability of GPFS when applied to even the most powerful computing systems," said Mark Seager of Lawrence Livermore National Laboratory. "The ability to leverage this file system architecture in new ways and across various systems will provide a new value to those employing high performance computing today -- and will open doors to a whole new class of applications."
The IBM software, called General Parallel File System, is an advanced file system for high performance computing clusters that provides high speed file access to applications executing on multiple nodes of a Linux or AIX cluster. General Parallel File System, scalability and performance are designed to meet the needs of data-intensive applications such as engineering design, digital media and entertainment, data mining, financial analysis, seismic data processing and scientific research.
Scientists at the forefront of supercomputing have used General Parallel File System, to create a scalable parallel file system that is capable of supporting hundreds of terabytes of storage within a single highly reliable file system.
IBM has several initiatives created to encourage General Parallel File System, support on all hardware platforms, including non-IBM hardware. IBM is offering access to General Parallel File System source code to eligible clients, who may choose to adapt General Parallel File System, to other platforms and share their work with other licensees, further promoting collaboration and innovation in file systems. By providing access to the source code for General Parallel File System, IBM is continuing its commitment to building the robust ecosystem of products that partners have come to expect when working with IBM.
About Lawrence Livermore National Laboratory
Founded in 1952, Lawrence Livermore National Laboratory has a mission to ensure national security and to apply science and technology to the important issues of our time. Lawrence Livermore National Laboratory is managed by the University of California for the U.S. Department of Energy's National Nuclear Security Administration.
System i, System p, System x, System z, BladeCenter, and Supercomputers
1 According to the November 2005 TOP500 List of Supercomputers, found at www.top500.org.