News Stay informed about the latest enterprise technology news and product updates.

Virtualization firm pushes high-performance computing towards the cloud

ScaleMP has made a business of turning standard servers into supercomputers. Will increased elasticity and on-demand functionality make high-performance computing cloud-capable?

The Cupertino, CA-based server virtualization firm ScaleMP claims that adding elasticity to its "Versatile SMP" (symmetric multiprocessing) product has made it cloud-ready. vSMP foregoes familiar virtualization technologies that carve a server into multiple virtual machines (VMs) by choosing to aggregate separate servers into one large VM.

More on high-performance computing and the cloud:
Medical researcher taps HPC on-demand service over Amazon EC2

HP intros RAM- and CPU- packed blade for cloud computing, HPC

Bandwidth issues deter HPC users from cloud services

vSMP Foundation for Cloud lets users deploy vSMP in a more flexible fashion, thanks to a change in licensing that allows new instances to fire up on nearby networked hardware clusters instead of site installations. It's a far cry from using Amazon Web Services (AWS) to build your own supercomputer, but it is a step towards on-demand, elastic computing in the conservative world of high-performance computing (HPC).

"For us, the trick is, it's all used on a network key…it eliminates the hardware lock [license token]," said Greg Keller, technical principal at R-Systems, a company that provides business-process outsourcing and software development services. Keller said he runs about 1,400 commodity servers, and he uses ScaleMP to aggregate them into groups for more firepower.

R-Systems found vSMP Foundation especially attractive because it doesn't require you to buy licenses until they're needed, said Keller. He pointed out that ScaleMP lets him build very powerful machines out of commodity hardware, and vSMP Foundation for Cloud was another way for ScaleMP to target users without the resources to build a true grid platform. He called it a way out of the "hug you closer syndrome" endemic to high-end computing hardware providers.

The other advantages, said Keller, involve operations costs and the ability to easily decommission unused clusters to save on power and maintenance.

"The smaller customers want to expand without expense," said ScaleMP founder and president Shai Fultheim.

ScaleMP customers could equal the capabilities of high-end, multi-core systems like the HP Superdome on commodity servers for a fraction of the cost, especially with the emergence of the Intel Nehalem and other rapidly proliferating multi-core CPUs, Fultheim said.

The smaller customers want to expand without expense.
Shai Fultheim, ScaleMP founder and president,
vSMP can equal the Superdome's actual cores by aggregating up to 128 cores in one VM. That could be had for well under $100K in hardware, according to Fultheim, while Superdome pricing starts in the millions.

ScaleMP plans to pick up some of that extra cash its customers won't be spending on a supercomputer. The vSMP Foundation for Cloud is $65,000 for an initial license and $650 for every additional machine license used, which is a steep increase over its base vSMP product, which starts at $37,000.

ScaleMP will also have an uphill battle against clustering technologies, such as the open source Nimbus and the commercial Platform Computing, both of which let users mimic HPC clusters on public resources and evade hardware and software investments.

Carl Brooks is the Technology Writer at Contact him at

Dig Deeper on High-performance computing in the cloud

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.