NPACI Rocks Marks Milestones: 100 Systems, 5,000 Processors, 15 Teraflops

Researchers with the National Partnership for Advanced Computational Infrastructure (NPACI) at the San Diego Supercomputer Center (SDSC) marked significant milestones this week for the NPACI Rocks toolkit, which enables users to more easily set up and manage powerful yet inexpensive cluster computers. With at least 100 systems created using the package, the processing potential of computers administered by the NPACI Rocks software suite now exceeds 15 teraflops -- computing power that researchers all over the world are using to advance science in fields from biomedicine to geophysics. "The aggregate power of the clusters running NPACI Rocks is in the same league as the largest supercomputer systems in the world," said Philip Papadopoulos, program director for SDSC's Grid and Cluster Computing group. "The number of systems and processors demonstrates the wide acceptance of NPACI Rocks in the user community." "I am extremely happy with the performance of the NPACI Rocks cluster distribution," said Tim Carlson of Pacific Northwest National Laboratory in Richland, Washington. "I have been installing clusters for the past five years and have tried many types of software. In my experience, Rocks clusters provide the right combination of stability, maintainability, and ease of installation." According to the Rocks Register at http://www.rocksclusters.org/rocks-register/, 100 computer systems -- many with true supercomputer processing power -- have been created using NPACI Rocks, and more than 5,000 CPUs run the software suite. A Pentium 4 cluster used in the Bio-X project at Stanford University runs at more than 3.3 teraflops and five clusters at Brown University have a total processing power of 1.2 teraflops. Cluster computers at the Karlsruhe Research Center in Germany (1.1 teraflops), the University of South Carolina (671 gigaflops), the University of Nevada (614 gigaflops), and Scripps Institution of Oceanography (355 gigaflops) also stand out as powerful systems for scientific research. The latest version of the toolkit, released a month ago, works both with 32-bit CPUs from Intel, AMD, and other chip makers and with 64-bit systems based on the Intel Itanium2 processor. (The Itanium2 also is used in "off-the-shelf" supercomputers from such vendors as IBM and Hewlett-Packard.) "The clusters listed on the register illustrate the scalability and versatility of NPACI Rocks," said Mason Katz, group leader for the software development effort. "At one end of the scale there's a five-processor Pentium II system with only 1.5 gigaflops of power. There are all kinds of 32-bit and 64-bit configurations all the way up to Stanford's 604-processor Bio-X cluster at 3.3 teraflops." Commodity clusters based on PC-type processors ("Beowulf clusters") provide impressive power for the cost of their hardware. But the cost of managing them -- ensuring that all of the nodes of a system have a consistent set of software when patches and new versions of the operating system, utilities, and tools are released -- can be a real burden to system administrators. Unfortunately, the costs of not managing a cluster can be even more expensive, if security holes and known software bugs remain unpatched because the system software isn't upgraded often enough. "SDSC's Cluster Group and UC Berkeley's Millennium group began to work together on on the then newly-formed NPACI clusters project exactly three years ago this month," Papadopoulos said. NPACI Rocks was the outgrowth of this collaboration, and the first public release was six months later at the SC2000 conference. "Over these three years, our goal has been to make clusters easy to deploy, manage, upgrade, and scale." "Typical NPACI Rocks users aren't cluster experts, they're smart people who need clusters as tools to do their jobs," said Greg Bruno, a researcher in the group. "A scientist who uses a Beowulf cluster shouldn't have to take up a second career in system administration. That's why we've automated NPACI Rocks to the point where only a handful of simple steps is required to bring up a full-featured cluster." "The Rocks Register is completely voluntary," Papadopoulos said, "and we know that we have more users and more processors administered by our software than the list implies. In fact, we'd like to encourage people out there who are running NPACI Rocks to sign up, just so we can gauge where our development efforts would be best spent." "We tried several clustering alternatives before settling on Rocks as our default system two years ago," said computational chemist Frederick P. Arnold, Jr. of Northwestern University, who manages a 64-bit Itanium2 system and two 32-bit clusters with NPACI Rocks. "It has proven easy to install, configure, extend, and use. It is extremely robust in production, and now forms the core computing environment for Northwestern Chemistry's Theory Group. It also evangelizes well to other groups once they see it in operation." The Linux Competency Centre at Singapore Computer Systems (SCS-LCC) has set up a new 60-processor Itanium2 cluster for the Singapore-MIT Alliance (SMA) at the National University of Singapore to support projects ranging from computational fluid dynamics to bio-engineering. About 50 SMA researchers and post-graduate students use the system. "The team took less than a day to install the cluster with Rocks and getting the cluster operational," said Laurence Liew, manager of the SCS Linux Competency Centre. "This is a testimony to the amount of work that has gone into making Rocks one of the best and easiest to use cluster toolkits in the world." NPACI Rocks is attractive to system administrators not only because it easy to use, but also because the terms of its open-source Berkeley Software Distribution (BSD) license mean that it's available to many users without charge. "In the past several months the use of NPACI Rocks has really accelerated," said SDSC software engineer Federico Sacerdoti. "We've heard that some big systems will be announced in the near future, and I wouldn't be at all surprised if the register hit an aggregate of 25 or 30 teraflops by the end of this summer." NPACI Rocks is developed by SDSC and partners at the University of California, Berkeley, Singapore Computing Systems, and individual open-source software developers. For more information on NPACI Rocks, see http://rocks.npaci.edu/. The NPACI Rocks Register is at http://www.rocksclusters.org/rocks-register/. --Mike Gannis