Cloud Infrastructure

Cloud infrastructure is the set of resources that workloads run on — compute, storage, networks, and the identity system that decides who may use them — delivered so that they can be obtained, scaled, and released on demand rather than bought and installed.

The most widely cited formal definition comes from NIST’s Definition of Cloud Computing (Special Publication 800-145, 2011), which describes a cloud infrastructure as the collection of hardware and software that enables the five essential characteristics of cloud computing. It splits that collection into two layers: a physical layer of server, storage, and network components, and an abstraction layer of software deployed across it, which is what actually delivers the cloud behavior. NIST’s physical layer lists servers, storage, and networks; identity is added here because in practice no resource can be used without it.

Those five characteristics are what distinguish cloud infrastructure from a rented rack. Consumers provision capabilities themselves without a request to the provider; capabilities are reachable over the network through standard mechanisms; the provider’s resources are pooled across many consumers; capacity can be provisioned and released rapidly, sometimes automatically; and usage is metered. Remove self-service or metering, and what remains is hosting.

The practical consequence is a boundary of responsibility, and it moves with the service model. In NIST’s description of infrastructure as a service, the consumer does not manage the underlying cloud infrastructure but does control operating systems, storage, and deployed applications. With software as a service, even application capabilities are outside the consumer’s control. Every step up that ladder trades control for work the provider takes on — which is why “we’re on the cloud” says little until someone states which layers the organization still operates.

It is also worth keeping apart from data infrastructure. Cloud infrastructure is general-purpose: the same compute, storage, and network serve a web service, a batch job, or a database. Data infrastructure is the part built on top of it and specialized for moving, storing, and serving data — table formats, catalogs, query engines — and the two usually have different owners.

The questions that matter in design follow from all this: which layers your teams operate and which the provider does, what happens to a workload when one pooled resource fails, and which identity is allowed to do what. How this layer relates to data platforms and developer platforms built above it is worked through in Data Platforms and Developer Platforms.