Cloud providers can offer services on top of the management layer, allowing customers to use cloud-based infrastructure in place of physical hardware such as Web servers or disk arrays. Amazon Web Services’ Simple Storage Service (S3), for example, allows customers to store and retrieve data through a simple Web interface, paying about 15 cents per gigabyte per month in the United States (with some additional charges for data transfers). The Elastic Compute Cloud (EC2), also from Amazon, provides virtual computers that customers can use for processing tasks. Prices range from 10 cents per hour to $1.25 per hour, depending on the size of the virtual computer and the software installed on it.
Beyond infrastructure offerings, however, companies are also providing more sophisticated services, including databases for managing information and virtual machines that can host applications written in high-level languages such as Python and Java, all of which can help developers get a new service or application to market faster. Google’s App Engine, for example, gives customers access to the technologies underlying Google’s own Web-based applications, including its file system and its data storage technology, Bigtable. Even when these services don’t use a layer of virtual servers (App Engine does not), they still allow users to expand and contract their usage with the flexibility that is the hallmark of cloud computing.
Perched on top of all these layers are the end-user applications, such as online calendars or programs for editing and sharing photos. By encouraging content sharing and loosening the limits imposed by our computers’ local processing abilities, these applications are changing the way we use software. While some–such as Web mail–predate clouds, building such services on clouds can make them more appealing says Rick Treitman, entrepreneur in residence at Adobe Systems and a driving force behind the Acrobat.com suite of applications (which do their computations on a user’s computer but draw data from a cloud as needed). For consumers, Treitman says, what’s most attractive about cloud applications is their constant availability, regardless of the user’s operating system or location, and the ease with which multiple users can share data and work together. But he notes that these qualities can come into conflict: allowing offline access to data stored in cloud applications, for example, offers a convenience to users but can create problems if multiple users access a document, change it offline, and then try to synchronize their efforts. (For more about some of the technical challenges facing cloud computing, see “The Standards Question,” p. 59.) While Amazon and other providers make cloud services publicly available, some companies are turning to cloud-computing technologies inside their own private data centers, with the goal of using hardware more efficiently and cutting down on administrative overhead. And once a company sets up its own private cloud, it has a chance to take advantage of additional flexibility. For example, a specialty of Cohen’s company, Enomaly, is setting up overflow computing, also known as cloud bursting. A company can host its Web services and applications in its own data centers most of the time, but when a spike in traffic comes along, it can turn to outside providers for supplemental resources instead of turning customers away.
Ultimately, clouds could even change the way engineers design the computers that are increasingly embedded in everyday objects such as cars and washing machines. If these low-powered systems can reach out and draw any amount of computing power as needed, then the sky’s the limit for what they might do.