Evaluating the Role of Cloud Computing for Scientific Discovery. We will assess the suitability of the existing major cloud APIs by Amazon, Microsoft, and Google for scientific research and, as appropriate, create extensions to such that would improve their suitability for scientific discovery. We will work with these companies to get any extensions incorporated back into their APIs. We will investigate how to best exploit the cloud for scientific data and will demonstrate by making available petascale biology and climate datasets through open cloud interfaces and integrate them with major cloud providers (and others). The primary objective of this work will be to gain specific knowledge on how to appropriately leverage the tools, approaches, and platforms provided by cloud computing to augment ORNL’s existing computational platforms and better address the computational needs of ORNL and the Department of Energy.
Significance
Cloud computing is an emerging paradigm for provisioning and use of computing services. We ignore the (tremendous) hype around the term and focus on a simple definition based on its original conception, which is essentially computing as a utility. From the user’s viewpoint, cloud computing is elastic, meaning that the resources allocated to the user readily grow and shrink as needed, and the user pays only for the resources they actually use. Cloud computing is generally viewed as being location-independent, meaning that the user neither knows, nor should they care, the physical location of the resources used to satisfy their request. From the resource provider’s viewpoint, the user buys a certain amount (or quality) of service in a way that is divorced from the actual hardware used to provision it. This allows the provider much more flexibility for how best to satisfy the collective needs of their user base.
Cloud computing is already very much on the radar of the HPC community and various sponsors and the potential to become the next disruptive technology. DOE, DoD, and others have already expressed interest in better understanding the role of cloud computing within their overall computing needs, particularly at the mid-range and in rapid-deployment situations. There are also indications that DOE may wish to pursue the cloud to support computational biology. Therefore, it is of vital strategic importance for ORNL to develop experience and expertise in cloud computing so that we can compete effectively with other laboratories in future DOE activities in this area.
In this project, we will focus on some of the central questions of how to effectively deploy and use cloud for mid-range computing, and strategies for integrating cloud computing with more traditional computing facilities. The practical hands-on experience developed through this project will develop ORNL expertise in cloud computing and also solidify our interactions with Amazon, Microsoft, and Google. This project will allow us to explore the feasibility of strategic partnerships to better respond to our sponsors’ cloud computing needs.
Project Timeline