Can cloud computing be used effectively to measure the expansion of the universe? That’s the question the U.S. Department of Energy tackles in a recent report linked to its Magellan project.
It seems that at least for now, the Energy Department is better off using its high-performance computing (HPC) systems rather than cloud computing. The former systems are cheaper and offer better performance. Yet when it comes to flexibility, many scientists in the Magellan project seem to prefer the cloud.
About two years ago, the Energy Department began a research project to determine whether cloud computing can be used for the CPU-intensive processing tasks scientists do in national laboratories.
The key results of that research, which has just been published, boil down to:
• Cloud computing has both advantages and disadvantages for Energy Department scientists.
• The performance and scalability of cloud computing are not yet ideal.
• Commercial cloud services are in general, between 7 and 13 times more expensive than HPC systems in the scientific laboratories.
• If the Energy Department would move its NERSC computing center to the cloud computing services offered by Amazon, it would have to pay about $200 million yearly, which greatly exceeds its budget.
• Cloud computing facilitates user access to software, and makes control over the software environment easier.
Scientists surveyed during the research have pointed that the cloud enables them to share the setup of software and experiments with colleagues, which is highly desirable when it comes to large scientific projects that involve the collaboration of many scientists from different locations.
But perhaps the most important find of the survey – especially for advocates of cloud computing – is this: almost 40% of respondents think the cloud’s features attractive, although they understand at the moment their systems offer better performance.