Adaptive Enhances HPC Scalability, Integration

For companies struggling with the complexity involved in using high-performance computing tools, Adaptive Computing is offering new releases – version 7.0 – of its Moab Cloud and Moab HPC product suites. The enhanced scalability and ease of integration, the company hopes, will enable more organizations to “gain the benefits of HPC,” and “expand the market,” especially as they move to the cloud, says Adaptive CEO Robert Clyde.

“You don’t generally put high-performance computing and cloud in the same place,” observes Sam Barnett, directing analyst for data center and cloud, at Infonetics. “Generally speaking, high-performance computing is a very, very small market.” He says HPC is often thought of in the same space as super computers, but that “the age of the supercomputer fell off the earth many years ago,” and only a few are sold every year, mainly to the government for military purposes, energy research and other areas that require massive computational power.

The challenge in super computing is how to effectively take a task, divide it up among different processors and make sure they provide whatever data is needed. “That’s always a problem when building these massively parallel systems,” says Barnett, adding that Adaptive appears to be addressing some of those challenges. “Now we’re doing super computing functions through massively parallel systems. That’s where Adaptive’s tools can come in handy. You can be managing hundreds of thousands of individual processor nodes.”

The HPC industryturned in a strong showing in 2010, surging 22.4% to $25.6 billion in total product and services revenue, and performance is expected to continue to grow at a 7% compound annual growth rate, reaching $36 billion in 2015 (Intersect360 Research Worldwide). According to IDC, HP accounted for 32% of 2010 HPC revenues, just edging out IBM with 30%. Dell held down third place (15%), followed by Cray and SGI (3% each) and a host of other vendors, including Hitachi, Fujitsu and NEC. For 2012 IDC just forecast that HPC server revenue will reach a record level of $10.6 billion, en route to $13.4 billion in 2015.

Moab 7.0 HPC Suite manages batch workloads, primarily for technical computing applications such as simulations, says Clyde. “Most workloads managed by Moab HPC Suite run on physical, non-virtualized machines.” Moab HPC Suite contains components that are characteristic of high-performance computing, such as the Torque resource manager which starts, stops and monitors batch jobs, he says.

“It looks like what they’re offering is a resource management tool geared toward improving service delivery,” says Barnett. “What that means in very complex computing delivery environments, is mistakes can be very, very costly in terms of configuration. It looks like their HPC suite product is really designed to eliminate some of the problems associated with that in configuration management, and is designed to boost overall productivity and hardware utilization through a virtual machine manager.”

Barnett believes this will help advance the HPC market. “Anything they can do to solve problems in the supercomputing world will be welcome to the high-performance computing community.”

Moab 7.0 Cloud Suite manages private clouds, primarily for business computing applications including web servers and application servers, Clyde says. Most workloads managed by Moab Cloud Suite run on virtual machines, and the Moab Cloud Suite contains some components that are unique to private cloud environments, such as the service catalog and the cloud operations database, he says. The product can also manage physical provisioning.

“Moab HPC Suite focuses on physically-provisioned machines, since most HPC users do not want the performance penalties associated with virtualization,” says Clyde, noting that the HPC and cloud markets are merging. He says Adaptive may or may not introduce a specific virtualization edition at some point.

The industry trend toward deploying private clouds is growing as enterprises see the cost and staff time reductions they produce, he says. That means cloud platforms have to become easier to use to meet organizations’ needs. The Moab Cloud Suite 7.0 release is the company’s first to include components like provisioning managers, monitoring systems and cloud operations databases out of the box, he says. In this space, they compete with BMC, he adds.

Barnett says there are only a few cloud service providers offering high-performance computing capabilities in a cloud construct, including Sabalcore Computing Inc., which offers an HPC on-demand service in the cloud. According to Clyde, in the HPC space Adaptive competes primarily with Altair and Platform.

Learn more about Strategy: Hadoop and Big Data by subscribing to Network Computing Pro Reports (free, registration required).