Feeds

Double-Take helps Microsoft Flex HPC muscles

Diskless iSCSI booties

Next gen security for virtualised datacentres

Microsoft is trying to wiggle its way into HPC shops with its Windows HPC Server 2008 variant, which includes similar tools that a Linux distro uses to support parallel supercomputing workloads.

Redmond apparently sees a nascent opportunity in the HPC space. Although the jury is still out about whether or not supercomputer users will ever sacrifice some computing power and memory to run hypervisors to carve up their cluster server nodes, HPC shops definitely want to be able to change operating systems and software stacks on their hundreds or thousands of server nodes more quickly than they can with sneakernet and grad students.

And that's the desire Microsoft wants to exploit — and it looks like the R2 update of its HPC Server, which is still in beta and which is expected later this year, will bring HPC Server 2008 to something close to parity to a Linux HPC stack as well as being able to support CPU-cycle harvesting from Windows 7-based PCs on the network. But to take on Linux in the HPC racket will require lots more capabilities.

So Microsoft approached Double-Take, one of its disaster-recovery software partners (which is in the process of being eaten by high-availability software vendor Vision Solutions for $242m), to adapt its Flex bare-metal provisioning and remote boot tools so it could support diskless provisioning of compute nodes over iSCSI links back to storage arrays for Windows HPC Server 2008. HPC shops are notorious cheapskates, so any piece of iron not in a server means less money spent and less heat, which in turn means more flops to do real work.

The resulting product, called Flex for HPC, can provision 100 diskless HPC nodes in under five minutes, according to Steve Marfisi, product manager for the Flex product line at Double-Take. It takes longer to do a larger number, and initial tests have shown it takes 90 minutes to juggle the images and fire them up for around 500 nodes. The reason is that the Windows cluster doesn't just boot from a raw OS image, but also has what Marfisi calls "differing disks", which Flex already had for its Windows replication software, so each node can not only get the raw image, but have its own unique software and data. This personalization was created for the Flex virtual desktop provisioning tool, where PC images are booted off the network but allow for each user to have some customization.

Flex for HPC only works in conjunction with Windows HPC Server 2008 R2 because it needs the iSCSI provider that Microsoft put into the R2 stack; the prior HPC Server 2003 and 2008 editions did not have this iSCSI provider. The tool also makes heavy use of the APIs in the cluster head node to control when what compute node gets updated, which Marfisi says has a sophisticated wizard that lets you "step back and get a coffee". Flex for HPC costs $1,995 for 40 compute nodes, plus $50 for each additional node after that.

What about Linux HPC nodes? Well, the existing Double-Take Flex product could be used to provide diskless Linux booting for each node on a cluster, but it doesn't have those head node wizards that automate the booting of compute nodes that Windows HPC Server 2008 R2 has and uses in conjunction with Flex for HPC. Wouldn't it be funny if you could make HPC Server 2008 R2 do remote booting of Linux images? We'll see if someone does a hack for that. ®

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
Death by 1,000 cuts: Mainstream storage array suppliers are bleeding
Cloud, all-flash kit, object storage slicing away at titans of storage
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
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
VMware vaporises vCHS hybrid cloud service
AnD yEt mOre cRazy cAps to dEal wIth
prev story

Whitepapers

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.
7 Elements of Radically Simple OS Migration
Avoid the typical headaches of OS migration during your next project by learning about 7 elements of radically simple OS migration.
BYOD's dark side: Data protection
An endpoint data protection solution that adds value to the user and the organization so it can protect itself from data loss as well as leverage corporate data.
Consolidation: The Foundation for IT Business Transformation
In this whitepaper learn how effective consolidation of IT and business resources can enable multiple, meaningful business benefits.
High Performance for All
While HPC is not new, it has traditionally been seen as a specialist area – is it now geared up to meet more mainstream requirements?