CPlant: The Largest Linux Cluster
With only half of it operational at the time of this writing, the Cplant(TM)/Ross system is already ranked 30th among the world's most powerful supercomputers, and is the largest Linux cluster in existence. Cplant is the result of a Sandia National Laboratories project started five years ago. Its story illustrates the possibilities of systems constructed in-house from commodity, mass-market components, and its construction is the triumph of an idea carried out with foresight by a handful of dedicated engineers.
Towns, mayors, or divide and conquer: The Chiba City I Scalable Management Architecture
Chiba City I is a 314-node Linux cluster in the Mathematics and Computer Science Division of Argonne National Laboratory. This article describes a delegation-based cluster management topology used in Chiba City. That topology consists of nodes available to applications (user nodes), dedicated management nodes (mayors), and a master management node (president).
Scale up your monitoring with Supermon
Supermon is a set of tools for high speed, scalable cluster monitoring. It monitors node behavior much faster than other common methods methods, such as rstatd, do. On a Pentium III/800 system, for example, Supermon can extract self-describing data from the Linux kernel at up to 6,000 samples per second.
Cluster Management with GNU cfengine
Cfengine is a cluster management system based on best-of-breed research and experience. It defines a set of principles for the configuration and maintenance of distributed systems. cfengine runs on hundreds of thousands of Unix and NT hosts around the world, some in the largest and best-known companies and organizations.
Large-scale clusters research - First HEPiX Large-Scale Cluster Computing Workshop
This article reports on the work of the Large Scale Cluster Computing Workshop, held at Fermi National Accelerator Laboratory.
Parallel Virtual Machine - Tuning PVM 3.4 for Large Clusters
The article discusses the trade-offs between MPI and PVM, and gives guidelines for improving PVM message transmitting performance to approximate that of vendor-optimized MPI implementations.
Farms, clones, partitions, packs, RACS, and RAPS - Improve your scalability vocabulary
This short paper tries to lay out a simple and consistent set of terms and some of the basic design issues of building huge servers from arrays of commodity components.
High Performance Mass Storage and Parallel I/O: Technologies and Applications
This article presents an overview of High Performance Mass Storage and Parallel I/O: Technologies and Applications, edited by Hai Jin, Toni Cortes, and Rajkumar Buyya, and published in 2002 by the IEEE Press and John Wiley and Sons. The book collects in a single volume 45 articles on the most innovative mass storage and I/O research from the last fourteen years. The first part of the review presents the reviewer's overall impressions of the volume, and gives an overview of the first part of the book's content.
