Cluster Basics

What is a Cluster?

A cluster is the aggregation of multiple stand-alone computers linked together by software and networking technologies to create a unified system.

Clusters are typically categorized into 2 general types:
  1. High Performance Computing (HPC), made up of markets traditionally serviced by supercomputers for applications requiring greater computational power than a single computer can provide; or
  2. Enterprise or High Availability (HA) with automatic failover, load balancing, redundancy, and other features that provide high reliability for the data center. Many HPC clusters also incorporate some of the features of HA clusters.

Application requirements vary between and within each of these system types. For this reason it's imperative that you choose a cluster partner that understands the intricacies of cluster design and can help you avoid the pitfalls of cluster deployment.

back to top

Cluster Architecture

Deployment of a highly efficient and usable cluster system requires more than simply linking servers together with high speed networks. Penguin Computing has introduced this value chain to demonstrate the complex nature of clustering and the manner in which we are turning the complexity into simple solutions through the use of our purpose designed hardware technologies, innovative Scyld clustering tools, effective partnerships, and industry leading expertise.

While other vendors are capable of delivering on value levels 1 and 2, the real magic happens at the higher levels. As a true cluster company, Penguin Computing focuses on all 4 value levels to deliver industry leading solutions and ensure that your cluster is optimized for application productivity.

back to top

From Beowulf to Production Quality Clusters

The original Linux cluster project, now widely known as Beowulf, was a computing breakthrough conceived at NASA Goddard in late 1993 by Donald Becker (CTO of Penguin Computing) and Thomas Sterling. They demonstrated that commodity clusters could do the work of multi-million dollar supercomputers, delivering the same results at a fraction of the cost. A more extensive history may be found on beowulf.org, the online community dedicated to the subject, or in Breaking New Ground: The Evolution of Linux Clustering. Today, Penguin is the leader in transforming the Beowulf concept into the delivery of commercially viable, production quality clustered computing solutions. We do this through finely tuned processes along with well-tested hardware, innovative software tools, and extraordinary customer service.

back to top

High Performance Computing (HPC) Clusters

Linux clusters are democratizing supercomputing for engineers, scientists, and researchers whose work demands the highest levels of computational analysis, modeling, and simulations. The Customers section of this site explains how innovative teams in a variety of industries are using HPC clusters to help speed up product development and groundbreaking research.

back to top

Dramatically Simple Clustering

Penguin Computing simplifies the cluster deployment process. Key to this simplicity is Scyld ClusterWare - powerful, elegantly simple Linux clustering software that makes a cluster appear and act like a single system. So whether you are a systems administrator or a scientist, Scyld ClusterWare dramatically simplifies clustering. Installation, administration, security and monitoring, are managed from a single point. With a powerful unified process space, you can deploy, manage, and run complex applications from a single point, making it truly as easy to use as expensive SMP environments.

Scyld ClusterWare is flexible and simple in design, so it?s extraordinarily scalable, with the capability of adding or deleting nodes instantly on demand?you can start small and let your cluster grow as your business grows. It?s robust, reliable, and integrates seamlessly with Penguin hardware to provide the most advanced and usable clusters available. And because it's commercially developed, supported, and backed by our deep Linux expertise, it lowers the risk of your initial investment as well as your total cost of ownership (TCO).

back to top

Summary of Cluster Benefits

The clustered system architecture is gaining in popularity for both high performance and high availability applications because of these compelling advantages:

  1. Linux provides cost advantages, greater flexibility for programming, fewer licensing requirements, faster response times to bug fixes, better security, and enhanced performance over proprietary operating system platforms.
  2. Standard commercially available hardware is not only lower in cost, but it?s where the latest technological advances are appearing first.
  3. The cost savings on both hardware and software means individual departments can afford their own supercomputers.
  4. Installations can start small and easily scale as demands increase.

Penguin increases the value of clustering by providing application-ready systems based on cluster optimized hardware platforms, well integrated cluster productivity software, and proven integration and delivery processes. We make the promise of dramatically lower TCO a reality.

back to top