Red Hat simplifies platform for Linux supercomputing
Red Hat on Thursday introduced what it called the first fully integrated, Linux-based, high-performance-computing platform, claiming to undercut a similar Windows-based system recently introduced by Microsoft.
The Red Hat HPC Solution is an update of a product of the same name that was introduced in November 2007. That product had more limited options for hardware and support, and was not available outside the US. The new version is the first to be available directly through the Red Hat Network and to be backed by Red Hat's own international customer-support services.
Red Hat claimed the system, which is priced from $249 (£141) per node, costs less than Microsoft's Windows HPC Server 2008, which also integrates clustering and management components.
HPC Server 2008 costs a one-time fee of $475 per node, but support for enterprise customers is only available via a three-year enterprise-assurance maintenance agreement, the pricing for which Microsoft has not publicly disclosed. Microsoft has said it will release support-pricing estimates via its website on 1 November.
Red Hat HPC Solution is Red Hat's attempt to take advantage of a growing market for supercomputing clusters, which are increasingly used by businesses for tasks requiring intensive computing power, such as data warehousing, line-of-business applications and transaction processing.
Linux has traditionally been used for scientific research, and supercomputers in that field are dominated by Linux: for instance, the top five systems on the Top500 supercomputer list are all based on various versions of the operating system. Such systems are traditionally assembled on an individual basis, with their configuration determined by the needs of the organisation.
The growing use of high-performance computing (HPC) in businesses has encouraged the likes of Microsoft and Red Hat to introduce systems that already include all necessary components for setting up and maintaining a cluster. Red Hat said its platform can be deployed in under an hour.
Red Hat is also using the system to gain some headway in the HPC market, where it lags behind other distributions, such as Novell's Suse Linux. Red Hat is the dominant commercial Linux distribution in other market sectors, such as enterprise servers.
The stack is based on Red Hat Enterprise Linux 5.2 integrated with Platform Open Cluster Stack (OCS) 5, Platform Computing's cluster software framework. The system includes device drivers, an installer, management tools, a resource-application monitor, interconnect support and a job scheduler called Platform Lava.
OCS is based on open-source cluster-management tool Kusu, which supports the Suse and Ubuntu Linux distributions, as well as Red Hat, according to Red Hat. It also includes several other open-source tools, such as the Nagios system monitor, the Cacti node and cluster monitor and the Ganglia workload monitor. The version of OCS used in Red Hat's earlier collaboration with Platform Computing was based on a different open-source project, Rocks, Red Hat said.
Novell does not currently offer a comparable offering around its Suse Linux distribution, but Suse Linux is used in many prominent supercomputing projects, including two of the top five supercomputers on the Top500 list.
IBM, also a major player in the Top500 list, in July introduced the IBM HPC Open Software Stack, with initial support for Red Hat Enterprise Linux 5.2 and planned support for Suse Linux Enterprise Server 10, as well as various clustering and management tools. This stack is available for free but support is only provided via an IBM web forum, with a limited, fee-based support offering covering only particular stack components.
In September, Microsoft and Cray introduced the Cray CX1 supercomputer, pre-installed with Windows HPC Server 2008, with prices starting at $25,000 and aimed at markets including financial services, aerospace, life sciences and universities.