During the last several decades, dramatic advances in computing power, storage, and networking technology have allowed the human race to generate, process, and share increasing amounts of information in dramatically new ways. As new applications of computing technology are developed and introduced, these applications are often used in ways that their designers never envisioned. New applications, in turn, lead to new demands for even more powerful computing infrastructure.

To meet these computing-infrastructure demands, system designers are constantly looking for new systemarchitectures and algorithms to process larger collections of data more quickly than is feasible with today.s systems. It is now possible to assemble very large, powerful systems consisting of many small, inexpensive commodity components because computers have become smaller and less expensive,disk drive capacity continues to increase, and networks have gotten faster. Such systems tend to be much less costly than a single, faster machine with comparable capabilities.
Building systems from large numbers of commodity components leads to some signi_cant challenges, however. Because many more computers can be put into a computer room today than was possible even a few years ago, electrical-power consumption, air-conditioning capacity, and equipment weight have all become important considerations for system designs. Software challenges also arise in this environment because writing software that can take full advantage of the aggregate computing power of many machines is far more dif_cult than writing software for a single, faster machine.
Recently, a number of commercial and academic organizations have built large systems from commodity computers, disks, and networks, and have created software to make this hardware easier to program and manage. These organizations have taken a variety of novel approaches to address the challenges outlined above. In some cases, these organizations have used their hardware and software to provide storage, computational, and data management services to their own internal users, or to provide these services to external customers for a fee. We refer to the hardware and software environment that implements this service-based environment as a cloud-computingenvironment. Because the term .cloud computing. is relatively new, there is not universal agreement on this de_nition. Some people use the terms gridcomputing, utility computing, or application service providersto describe the same storage, computation, and data-management ideas that constitute cloud computing.
Regardless of the exact de_nition used, numerous companies and research organizations are applying cloud computing concepts to their business or research problems including Google, Amazon, Yahoo, and numerous universities. This article provides an overview of some of the most popular cloudcomputing services and architectures in use today. We also describe potential applications for cloud computing and conclude by discussing areas for further research.