updated 02:44 pm EDT, Thu June 28, 2012
Compute Engine takes aim at Amazon EC2 service
Google has introduced its "infrastructure-as-a-service" offering during its second-day keynote at the I/O developer conference in San Francisco. Named Google Compute Engine, the product will give businesses and institutions access to the Google datacenter infrastructure to handle large-scale computing tasks, with the ability to scale thousands of Linux virtual machines, in a move that directly competes with Amazon's EC2 and Microsoft's Azure.
Claiming it delivers 50 percent more “compute per dollar” than other platforms, Google SVP of technical infrastructure Urs Hölzle demonstrated the service scalability by applying 600,000 cores onto a genome app for the Institute for Systems Biology, with the increase from 10,000 taking just a few seconds. The system was able to find multiple associations in the human genome every second, compared to an average of one association every ten minutes from a 1,000-core cluster.
Although it could serve as an extension to Google App Engine, the infrastructure service is more focused on major processing and data-management tasks rather than the website and mobile app backends that App Engine is geared for. The service will provide options for virtual machines with one, two, four or eight cores, with up to 3.75GB of RAM per core. Pricing starts at $0.145/hour for a single-core machine.
Google Compute Engine is available for use from today onwards, although currently only as a limited preview. The company has yet to announce a date for full availability.