HPCwire: NASA Looks to Move Science Apps Into the Cloud

Since 1986 - Covering the Fastest Computers
in the World and the People Who Run Them

Translation Disclaimer

Subscribe | Sign In

Visit additional Tabor Communication Publications

Leading HPC
Solution Providers

February 03, 2010

NASA Looks to Move Science Apps Into the Cloud

by Michael Feldman, HPCwire Editor

It seems only natural that the US space agency would be casting its eyes toward the clouds. Sure enough, NASA is now looking to cloud computing to optimize the operation of the agency's IT infrastructure for some of its science codes. Like many commercial businesses and government organizations, NASA is being asked to do more computing with fewer datacenter resources.

To that end, NASA has awarded a two-year, $600,000 deal to Parabon Computation as part of the agency's Small Business Innovation Research (SBIR) program. Parabon is being tasked to deliver a platform that lets NASA scientists and engineers develop and run modeling and simulation applications as a service via a standard Web browser. The initial target application will be climate modeling, but since the solution is general-purpose, any application that can benefit from distributed computation, scientific or otherwise, is fair game.

Reston, Va.-based Parabon has been in the grid computing business for over a decade, and like many grid vendors of that era has evolved its portfolio to straddle the grid and cloud computing space. The company sells grid middleware but also runs a brokered computation service for both businesses and non-commercial organizations.

The company's flagship product, the Frontier Grid Platform, is now in its fourth iteration. Frontier is especially well-suited for data-heavy applications, but the product can be applied to any application that fits nicely with the grid computing model, including compute-intensive codes and distributed command-and-control apps. The company has a number of commercial customers in pharma and finance, but its largest user base is in the US government, especially the DOD and intelligence community.

"It's a solution that lets an organization realize more return out of the investment in the infrastructure they've already paid for," says Steven Armentrout, Parabon's founder and chief executive officer. When you consider all the idle cycles of desktops and datacenter servers, many organizations still waste 80 to 90 percent of their compute capacity. Armentrout says Frontier is there to make the most of that unused capacity.

And even though the grid capability of the Frontier platform will be used to execute NASA's simulation and modeling applications on a distributed infrastructure, the SBIR work will enable Parabon to incorporate a project-centric browser interface atop Frontier's SaaS interface, the goal being to provide a lifecycle development environment for NASA codes. Using this platform, a developer would be able to define a project, build it, run the applications, maintain the code base, share the project with others, and so on. Users will be able to do everything from editing the source code to specifying the execution environment -- all within a browser.

The platform is meant to run inside an organization's firewall, and can be deployed on nearly any machine. Since Frontier can talk with all flavors of Linux, Mac OS X and Windows, while also supporting VMware-compatible virtual clients, it can tap into everything from desktops to supercomputers. In the case of NASA, this means that the science projects can not only be widely shared across users, but can also take advantage of a center's entire compute infrastructure at execution time.

At a practical level, that means that nearly anyone with a browser and the right access will be able to start using NASA codes. In the current model, sharing these codes necessitates someone building a tarball (along with build instructions) and shipping it off to a user to unwrap and deploy on their own infrastructure. Given the complexity of most science codes, this is a time-consuming and error-prone process. The user must also take into account compatibility between the local environment and the one the NASA code was built with.

Creating a browser-based, service-oriented model should eliminate these kinds of pain points. In the press release announcing the SBIR deal, Mike Seablom, who leads the Software Integration & Visualization Office at NASA's Goddard Space Flight Center, said the new platform "has the potential to revolutionize how scientific software is designed, developed, deployed and used."

By and large, the HPC community has been wary of service-oriented platforms and cloud computing in general because this model usually adds a virtualization layer to the software. Typically this translates into slower performance -- not a big deal for software development work and many other apps, but a deal-breaker for compute-intensive workloads. But according to Parabon's Armentrout, virtualization doesn't carry nearly the performance hit it used to. "The technology has really advanced, and you're getting close to native speeds with VMMs these days," he says.

NASA's climate models will be used as a proof of principle for the new service platform, but Armentrout expects it to be used throughout NASA. More generally, Armentrout sees a big demand emerging for this model for many science codes, especially with the company's bread and butter customers in the DOD. For example, the Army's Program Executive Office for Simulation, Training & Instrumentation (PEO STRI) has also expressed interest in getting its HPC simulation codes up in the browser. If Parabon can prove themselves with this platform for NASA, it could fuel a lot of interest throughout the HPC community.

"I think it's going to transform the way scientific computing is conducted," predicts Armentrout.