Feeds

Red Hat chases Redmond with HPC play

Hello? We supercompute too

Boost IT visibility and business value

As the largest supplier of commercialized Linux operating systems, it is a bit embarrassing for Red Hat that in the supercomputer market that gave Linux legs, other distributions own the high performance computing (HPC) space. That includes Novell's SUSE Linux Enterprise Server, but also a smattering of other Linuxes.

Enter, stage right, Microsoft with its Windows HPC Server 2008 distribution last week and a partnership with Cray to go after the baby supercomputer market, a bunch of other players like SiCortex using Gentoo Linux (not a variant of Debian), and you can see that Red Hat wants to ramp up its presence in HPC to go after what would seem like some easy money.

Today, Red Hat announced the global availability of its Red Hat HPC Solution, which is a mix of the current Enterprise Linux 5.2 from Red Hat and an OEMed version of Platform Computing's Open Cluster Stack 5, a set of open source cluster management tools that Platform, a pioneer in grid computing, created from its experience with its proprietary Load Service Facility (LSF) tool for managing jobs running on grids.

Last November, Red Hat and Platform launched an earlier edition of the Red Hat HPC stack, which was based on Platform's OCS, itself launched in July 2006 for RHEL and its CentOS clone and certified only Dell iron. This product was only available in the United States.

Platform's earlier OCS toolset for managing clusters was based on the open source project called Rocks, created by the San Diego Supercomputing Center and owned and administered by the University of California at San Diego. According to Gerry Riveros, product marketing manager at Red Hat, the current version of the OCS 5 stack being used by Red Hat in the HPC stack uses an open source cluster management tool called Kusu, which supports Red Hat, SUSE, and Ubuntu Linuxes.

Last fall, Red Hat and Platform gave the impression that the Red Hat HPC stack was using Rocks. But apparently, it was based on Kusu, a much younger program but one that was designed from scratch to be easier to use than Rocks. (No one at Red Hat or Platform said this as part of the announcement. Go figure).

Anyway, OCS also has an open source cluster workload management tool called Lava, created by Platform, that is based on LSF, and uses the Nagios system monitoring tool, the Cacti node and cluster monitoring tool, and the Ganglia workload monitor for the cluster and its node - all of which are open source.

We have a price

The real big difference with this iteration of the Red Hat HPC Solution - why can't we outlaw the word solution in IT, please? - is the price. By which I mean that Red Hat is actually talking about it, unlike back in November. The latest version of the Red Hat HPC stack costs $249 per server node in a cluster.

Considering that Platform charges $150 per node per year to support OCS when it sells the support contract itself, this is a pretty aggressive price, especially when you compare this to the $299 charge for a standard subscription to RHEL Desktop 5, which offers 12x5 business support, or the $799 fee for a standard subscription for RHEL 5.2 on a two-socket server. The Red Hat HPC Solution price is a lot lower than what Microsoft is charging per node for Windows HPC Server 2008, which costs $475 per node.

The way the OEM deal works, Red Hat is selling and supporting the complete RHEL 5.2 and OCS stack through its Red Hat Network. The company takes the tech support calls and kicks back a royalty to Platform. (How much? Riveros could not say. But it is probably not $150 per node. Maybe half that if Platform is lucky).

Red Hat is suddenly excited about HPC for the same reason that Microsoft is. Setting up a Linux cluster is a pain, and it requires the installation, tuning, testing, and troubleshooting of all the software elements in the stack. People want to have a set of software they can buy that sets up a cluster, or better still, buy pre-configured clusters - which are called bright clusters in the HPC industry - and not even go through the pain.

That was the idea behind Platform partnering with Dell to sell OCS on pre-configured servers using RHEL. And now, Red Hat wants to take back control of the stack and offer the whole shebang not only on Dell iron, but on any x64 iron. He who controls the code controls the customer and therefore the money.

What about Novell?

The wonder is why Novell doesn't have a SUSE HPC stack, especially since SUSE Linux Enterprise Server has a higher market share in the HPC sector than it does in the commercial market at large. Back in August 1995, HP and Novell partnered to offer a pre-configured HPC stack with all kinds of goodies added on top of SLES. IBM offers some discounted pricing on SLES licenses for HPC nodes. The exact discount amount was not announced earlier this summer, however, when IBM started doing this.

And in July, IBM announced its own set of tools, called the HPC Open Software Stack, including RHEL 5.2 and IBM's Advance Toolchain for Power Systems, Simple Linux Utility for Resource Management (SLURM), Extreme Cluster Administration Toolkit (xCAT), and a bunch of install scripts for automating cluster setup. The stack also has, for an additional fee, IBM's C/C++ and Fortran compilers and related libraries.

IBM will be adding support for SLES, the Torque resource manager, the Maui cluster scheduler, and the Ganglia monitoring tool. And for a free, ut will weave in its own General Parallel File System or Intel's C and Fortran compilers and math libraries. The repository of open source code used in IBM's HPC stack is being maintained at the University of Illinois, and the software runs on IBM's Power and x64 rack and blade servers.

It is not yet supported on its PowerPC-based BlueGene/L massively parallel supers or its Cell-based blade servers. But here's the real kicker. IBM is offering the code for free, but it is offered as-is without commercial support.

Novell needs to get a proper HPC stack into the field - Platform is a longtime partner and almost certainly willing to help out - and IBM needs to offer support on its stack if they hope to compete with Red Hat and Microsoft. ®

The essential guide to IT transformation

More from The Register

next story
The Return of BSOD: Does ANYONE trust Microsoft patches?
Sysadmins, you're either fighting fires or seen as incompetents now
Microsoft: Azure isn't ready for biz-critical apps … yet
Microsoft will move its own IT to the cloud to avoid $200m server bill
US regulators OK sale of IBM's x86 server biz to Lenovo
Now all that remains is for gov't offices to ban the boxes
Flash could be CHEAPER than SAS DISK? Come off it, NetApp
Stats analysis reckons we'll hit that point in just three years
Oracle reveals 32-core, 10 BEEELLION-transistor SPARC M7
New chip scales to 1024 cores, 8192 threads 64 TB RAM, at speeds over 3.6GHz
Object storage bods Exablox: RAID is dead, baby. RAID is dead
Bring your own disks to its object appliances
Nimble's latest mutants GORGE themselves on unlucky forerunners
Crossing Sandy Bridges without stopping for breath
prev story

Whitepapers

5 things you didn’t know about cloud backup
IT departments are embracing cloud backup, but there’s a lot you need to know before choosing a service provider. Learn all the critical things you need to know.
Implementing global e-invoicing with guaranteed legal certainty
Explaining the role local tax compliance plays in successful supply chain management and e-business and how leading global brands are addressing this.
Build a business case: developing custom apps
Learn how to maximize the value of custom applications by accelerating and simplifying their development.
Rethinking backup and recovery in the modern data center
Combining intelligence, operational analytics, and automation to enable efficient, data-driven IT organizations using the HP ABR approach.
Next gen security for virtualised datacentres
Legacy security solutions are inefficient due to the architectural differences between physical and virtual environments.