Software deployment

When computers were extremely large, expensive, and bulky (mainframes and minicomputers), the software was often bundled together with the hardware by manufacturers. If business software needed to be installed on an existing computer, this might require an expensive, time-consuming visit by a systems architect or a consultant. For complex, on-premises installation of enterprise software today, this can still sometimes be the case.

However, with the development of mass market software for the new age of microcomputers in the 1980s came new forms of software distribution – first cartridges, then Compact Cassettes, then floppy disks, then (in the 1990s and later) optical media, the internet and flash drives. This meant that software deployment could be left to the customer. However, it was also increasingly recognized over time that configuration of the software by the customer was important and that this should ideally have a user-friendly interface (rather than, for example, requiring the customer to edit registry entries on Windows).

In pre-internet software deployments, deployments (and their closely related cousin, new software releases) were of necessity expensive, infrequent, bulky affairs. It is arguable therefore that the spread of the internet made end-to-end agile software development possible. Indeed, the advent of cloud computing and software as a service meant that software could be deployed to a large number of customers in minutes, over the internet. This also meant that typically, deployment schedules were now determined by the software supplier, not by the customers. Such flexibility led to the rise of continuous delivery as a viable option, especially for less risky web applications.