"Right now the climate models that we have are very complex, the software is upwards of 500,000 to 1 million lines of code," says Michael Seablom, head of the software integration and visualization office at NASA's Goddard Space Flight Center in Maryland.
A climate model might, for example, predict what would happen to global temperatures over the next hundred years if humans double carbon dioxide emissions.
"Trying to get the models running is difficult and it costs us a lot of money here because we have to help groups build the system on their local machine," Seablom says. "The problem with that is if you're a graduate student, you could spend months just trying to get the model running and verify that it's working correctly."
NASA's goal is to build a Web portal for investigators to log onto, allowing them to run the climate models on remote systems provided by the space agency. The grid computing software company Parabon Computation was awarded a two-year, $600,000 contract to help NASA build the system.
Parabon says the Web-based platform will be built upon its Frontier Grid software, which can take idle computing capacity from many machines and manage it as one large computational grid, with applications running on virtual machines.
Seablom says NASA's climate modeling teams will tap into processors from NASA's Nebula cloud computing platform and could someday purchase computing cycles from public cloud platforms.
"I hate to use the term 'cloud computing' because I've heard the term so much and I'm sick of it," Seablom says. "But the fact of the matter is this is a very good cloud computing model and we're going to save a lot of money doing it. I'm very excited."
Parabon uses its own software to buy idle computing capacity from universities and businesses and then resells computing cycles as an online service, says Parabon CEO Steven Armentrout. In this case, Parabon is selling its platform to NASA as an enterprise software package to be deployed internally behind the NASA firewall, he says.
"These tools -- such as a browser-based source code editor, online collaboration utilities, and virtualized build and runtime environment management interfaces -- will allow developers to more efficiently create and modify a wide variety of high-performance computing (HPC) applications," Parabon states in an announcement of the NASA contract.
Frontier can be used to harness the unused CPU power of desktops and servers, Armentrout notes.
"I believe they [NASA] have 80,000 desktops," he says. "If they were to put Frontier on all 80,000 they would have one of the fastest supercomputers in the world."
Although the system will initially run climate models, it can be used for many types of scientific research.
"If it's successful, and that's a big if, we could be buying our compute cycles over the network, as opposed to just having them in-house, or it could be a combination of the two," Seablom says. "The important thing for the taxpayer is we think this will save money."
Follow Jon Brodkin on Twitter: www.twitter.com/jbrodkin
This story, "NASA building cloud service for climate modeling" was originally published by Network World.