'Cloud' computing in simulation

DEFENCETODAY Clusters, grids and more recently, clouds, have demonstrated great utility and have considerable potential for handling large computational loads previously out of reach for single processor machines, including traditional supercomputers. This in turn has led to numerous predictions of a revolution in computing, and public claims that this technology will reduce the demand for skilled computer scientists, especially those with technical programming skills in operating systems, networks, and systems integration. The material reality of programming a distributed computer system with numerous processing cores, usually geographically scattered across multiple sites and linked through high speed network connections, is however quite different to the optimistic predictions of the non-technical advocates of this technology. In general, making software run reliably, robustly and quickly across a cluster, grid or cloud actually requires personnel with special skills. If applications are successfully developed or rehosted to a cluster, grid or cloud computing system the resulting performance gains can often be remarkable. If the application is readily parallelised and thus compatible with a distributed computing system, performance gains directly proportional to the number of computing cores can actually be achieved. To date this technology has produced its greatest impact in large scale scientific simulation work, spanning areas such as computational electromagnetics, computational aerodynamics, molecular biology, chemistry, pharmacology, geophysics, nuclear physics, astronomy and astrophysics. In recent years it has also penetrated into very large scale web based search and retail engines, with major players like Amazon. com, Google.com and Microsoft offering a cloud computing service product, based on their in-house distributed computing environments. As with any emerging or maturing technology with great potential, distributed computing, especially cloud computing, has stimulated a cacophony of commercial hype around the technology, and unprecedented numbers of trademark applications centred on names incorporating 'cloud' or 'clouding computing'. This should come as no surprise, as similar behaviour was behind the high technology stocks bubble, which burst around the turn of the millenium producing an unwarranted slump in the sector as wary investors steered away from any and all technology products. Given the failures in due diligence and the propensity to believe hype that preceded this event, there is every reason to expect another cycle of irrational investment behaviour as distributed computing technology matures and penetrates the market. The technology may be sound, but the marketing and sales around it is usually not. The idea of providing computing power as a 'utility' product in …