Recently, the Extreme Science and Engineering Discovery Environment (XSEDE) and the National Science Foundation Division of Advanced Cyberinfrastructure , released the results of the XSEDE Cloud Survey Report.
The report by David Lifka (Cornell University PI), Ian Foster (Argonne National Laboratory/University of Chicago), Susan Mehringer (Cornell), Manish Parashar (Rutgers University), Paul Redfern (Cornell), Craig Stewart (Indiana University), and Steve Tuecke (Argonne National Laboratory/University of Chicago) examined cloud computing use cases in research and education.
Three of the cloud benefits reported in the study struck me as particularly interesting: first is cost efficiency; secondly, the broader, more diverse adaption; and finally is the ability found to rapidly create prototypes.
Simply put, cloud computing offers a variety of alternatives to more traditional research computing, many of which are far more cost-effective. According to the study, some of the areas in which users found savings included:
Since application demand is often cyclical in research - there are peaks and periods of under utilization - the costs of building an internal compute infrastructure is increased because of the costs associated with "down time." The ability to "pay as you go" proved to be an important consideration for survey respondents when they weighed the advantages of purchasing on-premise hardware or computing in the cloud.
The study also noted that cloud computing may also add value through better planning of batch workloads, tracking usage trends, freeing up staff to focus on higher-order research needs, algorithm development, etc.
Broader, more diverse adoption:
According to the study's authors, "cloud appeals to a broad class of researchers, many of whom are not traditional HPC users." For example, one respondent noted that the organization's cloud solution is primarily meant to help domain scientists, who lack more advanced IT skills.
More researchers are beginning to see the value of advanced computation when cloud is incorporated as a part of a comprehensive cyber infrastructure portfolio. Researchers and educators pointed out an array of benefits to cloud usage, including:
Platform services such as storage and programming abstractions.
Outside of academia, the authors say barriers to entry for small to medium-sized businesses also may be decreasing. This finding mirrors HPC adoption within this market. As an example, the study offers The UberCloud Experiment, which is "exploring the benefits and challenges of accessing the cloud for CAE and other simulation applications that given additional compute resources could speed up product design or improve product quality." They also cite a report by the Council on Competitiveness (Make: An American Manufacturing Movement ) that suggests by providing "agile services that are accessible regardless of company size or location," cloud computing could be game-changing technology.
Rapid creation of prototype
One of the very tangible advantages of cloud infrastructure is the ability for departments without compute resources to try new ideas or classes of problems. Clouds allow researchers to forgo the need to deploy hardware. Computing can be conducted on more convenient time tables when it's no longer necessary to compete for access to on-premise or shared resources. This is especially true when these resources may be inundated with other priority projects. Very simply put, they provide research agility, allowing for "fastfalling" (quick testing of ideas), the possibility to do the unexpected, and an ability to create any number of simulated prototypes without incurring the cost of actual manufacturing.
Along with cost, survey respondents also noted cloud scalability as a benefit. Together these assets allow researchers to explore different classes of problems quickly, and enjoy rapid prototyping - both of which can lead to new avenues of research.
The scope and depth of this study accurately illustrates not only the advantages being realized by cloud computing, but the potential offered by cloud computing. The authors collected data from eighty international projects, and used a cross-section of cloud users representing 21 disciplines. Among the disciplines, science and engineering, the humanities, arts, and social sciences were employed. The authors also examined both quantitative dimensions and the qualitative experience.
You can read the survey in its entirety here.