News
Rocks Cluster Management Toolkit Rolls Over New Milestones
Published July 19, 2004
Aggregate power of Rocks clusters now exceeds 50 teraflops; version 3.2 adds new features
The Rocks development team at the San Diego Supercomputer Center (SDSC) is celebrating two milestones: passing the 50 teraflops mark in aggregate computing power, and the release of version 3.2 of the cluster computer management software suite.
"As of today there are 241 systems listed on the Rocks Register, with a combined power of more than 51 teraflops," said Mason Katz, group leader for the Rocks software development effort. "To give you an idea of how much performance that is, the Earth Simulator in Japan-the most powerful computer on the planet-is rated at 41 teraflops, 41 trillion mathematical operations per second."
Rocks is a user-friendly software package that gives the user of a cluster computer the ability to quickly and easily build the system software suite, install new software on the processors, and manage the system configuration. It provides a Linux cluster environment that enables users to start, monitor, and control processes on cluster nodes from the cluster's front-end computer while supporting standard Linux interfaces and tools.
The Rocks development group at SDSC has recently released version 3.2, which includes new features and enhances support for 64-bit Itanium™ processors, 64- and 32-bit AMD Opteron™ processors, and 32-bit only x86 CPUs.
Rocks clusters range from small systems based on only a couple of PCs to a supercomputer at Fermilab with more than 1,500 processors. Systems at Germany's Grid Computing Centre Karlsruhe, the Texas Advanced Computing Center, and Stanford University are rated at more than 3 teraflops apiece.
"Five of the machines on the current Top 500 List of high-performance computers are Rocks clusters," said Rocks team member Greg Bruno. "As it happens, the system that put the Rocks Register over the 50 teraflops mark was a new machine at the AMD Developer Center in Sunnyvale, California based on 256 AMD Opteron processors, with nearly a teraflop of power. We were pleased to see this because we've been working closely with AMD over the past several months to enhance Rocks' support for their 64-bit capable AMD Opteron processor line. In fact, our group currently is using an AMD Opteron processor-based cluster to help further our development efforts."
"Rocks has established itself as a standard open-source clustering software stack, and we are delighted to assist SDSC's development effort," said Pat Patla, director of enterprise and server/workstation at AMD. "Clusters based on Rocks and the AMD Opteron processor are a competitive, very capable choice for high-performance computing."
Rocks 3.2 builds on previous releases, and includes a number of application-specific extension modules. Rocks "Rolls" provide a mechanism for packaging operating system, middleware, or application software into the default cluster. System builders can configure clusters with the capabilities that they need, while side stepping unnecessary extensions.
"Essentially, Rolls capture what a competent cluster administrator would do to add new capabilities and then automate that for the non-expert," said Philip Papadopoulos, program director for SDSC's Grid and Cluster Computing group.
The Rocks 3.2 release includes updated Rolls and two additional new Rolls for Pentium™ and AMD Athlon™ based clusters. The "Area51 Roll" contains the tripwire and chkrootkit system integrity tools. The "Condor Roll" adds distributed high-throughput features from the Condor project. A number of other extension Rolls have been freely available since version 3.0, including Rolls for HPC, Grids (based on NMI R4), Java, Condor, Sun Grid Engine, and Ninf-G.
About Rocks
The Rocks suite is freely available for download at http://www.rocksclusters.org/. The software is released in three versions: for 32-bit x86 processors, Intel's 64-bit Itanium series, and AMD's 64-bit Opteron processors. All Rocks source code is available in a public CVS (Concurrent Versions System) repository at http://cvs.rocksclusters.org/. Rocks 3.2.0 is derived from Red Hat's publicly available source packages (SRPMS) used in portions of their Enterprise Linux 3.0 Product Line.
Rocks is developed and maintained by the Grid and Cluster Computing Group at SDSC and by partners at the University of California, Berkeley, Scalable Systems in Singapore, and a number of individual open-source software developers. Rocks development at SDSC is funded by the National Science Foundation.
About SDSC
The mission of the San Diego Supercomputer Center (SDSC) is to innovate, develop, and deploy technology to advance science. SDSC is involved in an extensive set of collaborations and activities at the intersection of technology and science whose purpose is to enable and facilitate the next generation of scientific advances. Founded in 1985 and primarily funded by the National Science Foundation, SDSC is an organized research unit of the University of California, San Diego. With a staff of more than 400 scientists, software developers, and support personnel, SDSC is an international leader in data management, grid and cluster computing, biosciences, geosciences, network research, and visualization. For more information, visit http://www.sdsc.edu/.
Contact: Ashley Wood, awood@sdsc.edu, (858) 534-8363