On-premises as a service is a way of buying computing and storage in which the machines sit in the customer's own building, or in a partner's data centre, but the customer pays for them by use, as if it were renting from a cloud. On-premises means on the customer's own premises. It is also called on-premises cloud, or consumption-based infrastructure.
Why would a company want this? There are two usual ways to get computing. A company can buy servers and storage, paying a large sum up front, and then own and manage them for about five years. Or it can rent from a public cloud, paying monthly for what it uses, while the machines stay in the provider's data centre. Some companies want the cloud's way of paying, but cannot or do not want to put their data in a public cloud. Their data may be too large to move cheaply. Rules may require it to stay in a particular place. Or the app may need to sit close to the people or machines that use it.
How does it work? The provider installs more equipment than the customer needs today, so that there is room to grow. The provider owns the equipment, and remains responsible for repairing and upgrading it. A meter in the software records how much the customer actually uses each month, such as terabytes of storage or numbers of virtual machines. The customer is billed on that number, often with an agreed minimum. When the usage comes close to the installed amount, the provider adds more equipment.
Consider a chain of diagnostic centres in Indore. It does 2,000 CT and MRI scans a week, and each scan creates hundreds of image files. Its storage needs grow by about 6 terabytes a month. It does not want to buy five years of storage in advance. It also cannot send every scan to a public cloud, because doctors open the images in the same building minutes after each scan, and the internet link is too slow for that. So it takes on-premises storage as a service. The provider installs 400 terabytes in the centre's server room. In April, the chain uses 180 terabytes, and pays for 180. In May, it uses 186, and pays for 186. When usage reaches about 320 terabytes, the provider will add another shelf of disks, and the bill will still follow use.
The equipment is usually ordinary enterprise storage systems and servers, run by software similar to private cloud software. It can also sit in a colocation data centre, where the customer rents space and power instead of using its own building. A company that runs some apps in a public cloud and others on its premises in this way has a hybrid cloud.
How is it different from leasing? A lease is a fixed monthly payment for fixed equipment. Here, the payment follows use, and the provider adds capacity as it is needed.