A cloud is a collection of computers, usually owned by a single party connected together such that users can lease access to a share of their combined power. Examples include Amazon’s Elastic Compute cloud, Google’s App Engine or IBM’s Enterprise Data Center.
Clouds share two main attributes with grid computing:
However, while grid computing is more focused on openness, collaboration and sharing between different sites over a prolonged period of time, cloud computing is designed more for short-term, on-demand provision of computing power and storage. As processing power and storage is sold as a service, cloud providers can easily charge users for the services they require. It has been argued that clouds and grids are based on different paradigms with clouds being based on virtualization of resources, and grids being based on the sharing of resources across boundaries.
Virtualization allows you to run ‘virtual computers’ which include applications and operating systems of your choice, on one single machine. Any number of virtual machines can run independently on the same physical machine.
Many clouds use virtual machines to schedule tasks or ‘jobs’. Users can submit their jobs as virtual images that include all the required software and tools, creating the ideal environment in which to work on the job.
Text adapted from Grids and clouds: the new computing
Read more at the GridBriefing from GridTalk.