The European Space Agency (ESA) plans to launch a new scanning satellite to map the stars in our galaxy in more precise form than ever before, and researchers supporting the project, called Gaia, are evaluating whether Amazon's cloud computing service can handle the mammoth processing task.
"We have these big processing problems in astronomy, and we have to look at cost-effective ways for meeting a processing problem," says William O'Mullane, the Gaia science operations development director at ESA.
With help from consultancy The Server Labs, ESA is now evaluating whether it can use Amazon's Elastic Computer Cloud (EC2) to process star data sent back from space.
Though the launch date for the project is still about two years away, scientists need to decide how to handle processing the years of data streams expected to flow intermittently at 5Mbps from the Gaia spacecraft.
The goal is to catalogue about one billion stars that move through the expanse of space with huge precision, which will likely mean many gigabytes of data each day. "All stars move and traverse the sky," O'Mullane says. "We will measure proper objects and do the space-based math."
ESA is now seeking to determine whether it would be feasible to rent about 4,000 application servers from Amazon to run the programs ESA has written for processing the star data. This would involve exporting the collected star data -- which the Gaia project plans to store in an Oracle database in a Madrid location -- to the Amazon cloud over the Internet.
"This frees up the servers we have," O'Mullane says. So far, early estimates suggest there could be a 50% cost savings over doing the job in-house.
Initial tests conducted with help from The Server Labs have shown the Amazon cloud to be a technically feasible option, though another round of evaluations is underway, according to Paul Parsons, CTO at The Server Labs.
Some in Gaia's upper management have expressed uneasiness about cloud computing, especially if the cloud turns out to be an Amazon EC2 data center outside of Europe, O'Mullane says.
"They haven't gotten the idea it's a global idea," O'Mullane says, noting Gaia upper management would prefer a "European cloud with things running in Europe." The issue remains a stumbling block to settling on an approach with Amazon, which has a data center in Ireland and one on the East Coast of the United States that are under consideration.
O'Mullane says the security provided in Amazon's cloud appears to be suitable for the proposed task, noting there are firewall and file-encryption systems and IPSec-based VPN access that could be used to establish a virtual private cloud on an isolated subnet.