Armbrust et al. (2009) provide a broad overview of the costs, benefits, and challenges of cloud computing. Although they do not focus specifically on scientific and engineering applications, they discuss several issues that appear in this study, such as batch processing of parallel processing applications, compute-intensive desktop applications, data transfer bottlenecks, and data licensing issues.
Cloud computing for computational science and engineering is a very young but increasingly active area, as evidenced by new workshops emerging in 2010 such as the first Workshop on Science Cloud Computing (ScienceCloud) (http://dsl.cs.uchicago.edu/ScienceCloud2010/) and Cloud Futures 2010: Advancing Research with Cloud Computing Workshop (Faculty Connection, 2007). Some early experience reports have begun to emerge. Hoffa et al. (2008) explored the use of cloud computing for executing a scientific workflow in the field of astronomy. Lauret and Keahy used cloud computing resources to quickly perform a preliminary analysis of a nuclear physics experiment in time to submit a conference paper (Heavy, 2009).