Cambridge Online Systems Limited

LINUX for HIGH PERFORMANCE COMPUTING

Linux Growth
In today's fast paced world, reducing the time it takes to run your highly computational code can give you the advantage over your competitors when developing innovative products or making a scientific breakthrough. It takes supercomputer type performance to give you the edge provided by many technical and commercial applications.

Linux is already over 10 years old and, with it now established as stable, efficient and reliable software, it is coming of age as the fastest growing operating system and as a credible alternative to UNIX and proprietary systems. Users are taking advantage of Linux running in clusters of industry standard servers to provide high performance for critical applications, while enjoying a lower total cost of ownership.

The introduction of industry standard platforms by HP and others widens the affordability of power computing to commercial as well as scientific users, from what was the preserve of an elite of high-funded technical users.

Linux Clusters for High Performance Computing
The Beowulf Project was started at NASA in 1994 to develop affordable, high performance clusters of Linux based systems. The success of the project can be seen in the take up of Linux today.

The Top 500 Supercomputer Sites listing (www.top500.org) provides a reliable basis for tracking trends in high performance computing worldwide. The analysis over the last five years shows Linux with "commodity" hardware clusters having a growing presence in the Top 500 list to the point where it is the dominant architecture. Equally interesting is the growth of high performance computing systems beyond their traditional home ground of research and education and in to industry and commerce, with HP seen as a leader in providing systems across all levels of performance.

Whatever your organisation's needs, it is likely that a configuration built by Cambridge Online from HP's range of industry standard and Open Source components will meet your requirements for performance and warranted support. Cambridge Online can offer:

High Performance and High Availability Linux clusters offering cost effective and powerful computing solutions
HP StorageWorks Scalable File Share (based on Lustre technology) enables bandwidth to be shared by distributing files in parallel across clusters of industry-standard server and storage components, reducing input/output (I/O) bottlenecks and scalable from dozens to thousands of Linux clients using TCP/IP
Clustering with OpenSSI (Single System Image, a HP supported project) gives single management and security domain, dynamic load balancing and process migration for explicitly parallel applications and high scalability across nodes.

The Right Solution for you from Cambridge Online
Cambridge Online has supplied information technology for over 25 years and customers within commercial and academic research organisations have all benefited from our expertise. A knowledgeable team of technical consultants has experience of configuring and building Linux, UNIX and proprietary systems for demanding high performance applications and can understand the requirements of computational based researchers.

Before designing a best fit solution, we will work with you to capture and define your computational needs:

Is your critical application floating point or integer calculation intensive and what are its memory and I/O bandwidth requirements? Which HP industry standard platform is the best match:
- AMD Opteron x86-64 HP Proliant
- Intel Xeon IA32 HP Proliant
- Intel Itanium II IA64 HP Integrity
Do you need to 'scale-up' for HPC with SMP (Symmetrical Multi Processing), very large single memory, one O.S. kernel and superior single job throughput, or 'scale-out' with clustering, distributed memory and throughput of many jobs?
What sort of interconnect is right for your clustering (e.g. Myrinet, InfiniBand or Quadrics) considering price and speed of I/O?
Which distribution of Linux is right for you, e.g. Red Hat, SUSE or Debian?

This consultancy is augmented by the Cambridge Online HPC Laboratory containing 32 and 64-bit systems for clustering, SAN and virtualised storage and networking, all configurable to meet your proof-of-concept and benchmarking needs.

We see that many companies and organisations are going through the same decision making process to assess whether Linux should be part of their infrastructure and the best way to deploy it. Cambridge Online has the experience and skills to support you with Linux systems and this has been recognised by HP choosing us as an elite HP Preferred Linux partner and supplier of High Performance Computing. We offer a single point of contact and expertise for:

High Performance Linux Clusters
Migration from UNIX or proprietary system to Linux
Introducing Linux systems in to your existing infrastructure
Training requirements to develop Linux expertise
Service and support to keep your operation running
Complete IT systems infrastructure (storage, networking, racks, power supply and cabling)
Programming and software tools
Effective management of Linux systems
Application availability and compatibility with Linux distributions

Contact Cambridge Online to discuss the right cost-effective High Performance Computing solution for you.

Click here for all News & Events