Cloud computing is a model for delivering computing resources—including processing, storage, networking, and applications—through network-accessible services. Users obtain configurable resources on demand rather than managing every underlying physical system. Access commonly occurs through the Internet, although private networks can also support cloud services. The model combines resource sharing, automated provisioning, and flexible capacity; it is not simply another name for remote hosting or online storage. (nvlpubs.nist.gov)
Essential characteristics
The United States National Institute of Standards and Technology (NIST) formalized a widely used definition in September 2011. It identifies five essential characteristics:
- On-demand self-service: users provision resources without requiring a provider employee to handle each request.
- Broad network access: services are accessible through standard network mechanisms and different client devices.
- Resource pooling: physical and virtual resources serve multiple consumers, with allocation changing according to demand.
- Rapid elasticity: capacity can expand and contract as workloads change.
- Measured service: resource consumption is monitored and reported at an appropriate level, such as processing time, storage volume, or active accounts.
Elasticity does not mean that physical capacity is unlimited. Rather, available resources may appear effectively unlimited to an individual consumer. Metering supports usage-based charging, but cloud services may also use subscriptions or other commercial arrangements. (nvlpubs.nist.gov)
Service models
Cloud services are commonly classified by the capability delivered and the division of operational responsibility.
Infrastructure as a service (IaaS) supplies processing, storage, and networking resources. Customers can install operating systems and applications while the provider manages the underlying infrastructure.
Platform as a service (PaaS) supplies an environment for deploying applications using supported development tools and runtimes. Customers control their applications without administering the underlying servers or operating systems.
Software as a service (SaaS) provides applications operated by the provider. Users access them through browsers or program interfaces, usually with limited configuration control. Cloud-based email illustrates this model. These categories distinguish management boundaries rather than the physical location of a particular application. (nvlpubs.nist.gov)
Serverless computing further abstracts resource provisioning and server administration. In function-based services, events trigger application code, while the platform handles execution and scaling. Servers still exist; “serverless” describes what developers do not directly manage. It is broader than functions as a service and can include managed backend capabilities. (glossary.cncf.io)
Deployment models
NIST distinguishes four deployment models. A public cloud is available for open use by the general public; “public” does not imply that customer data are publicly accessible. A private cloud serves one organization and may operate on or off its premises. A community cloud serves organizations with shared requirements. A hybrid cloud connects distinct cloud infrastructures through mechanisms enabling data or application portability. These models concern access and organization, independently of the service model. (nvlpubs.nist.gov)
Using several providers is often called multicloud. It is distinct from hybrid cloud: multiple public-cloud services do not necessarily constitute an integrated combination of different deployment types. Such arrangements raise questions about interoperability, identity management, and workload portability. (nvlpubs.nist.gov)
Technical architecture
Cloud infrastructure includes physical equipment housed in data centers, an abstraction and control layer, and a service layer. Virtualization allows hardware resources to be represented as independently manageable computing environments. A virtual machine provides an isolated software-defined system, commonly including its own guest operating system. Virtualization supports pooling, but it is an implementation technique rather than the definition of cloud computing itself. (nvlpubs.nist.gov)
Cloud platforms coordinate provisioning, monitoring, and resource allocation across these layers. Distributed computing supports services spanning multiple machines, while load balancing distributes requests across execution resources. Replication and geographic distribution can improve availability, although application architecture determines whether these capabilities actually protect a workload against failures. (nvlpubs.nist.gov)
Managed services also expose higher-level capabilities, such as databases, storage, analytics, and messaging. Applications interact with them through an application programming interface rather than administering all supporting hardware and software. (aws.amazon.com)
Uses and economic trade-offs
Cloud resources support application hosting, backup, development environments, and data processing. Analytics and machine learning workloads can acquire substantial computing capacity without purchasing equivalent hardware. Distributed processing frameworks such as MapReduce can run on cloud infrastructure, although communication and storage requirements influence performance. (aws.amazon.com)
The economic model shifts some expenditure from infrastructure acquisition toward service consumption. Shared infrastructure can improve utilization, while temporary allocation suits workloads with variable demand. These advantages do not guarantee lower total cost: utilization, administrative effort, network transfers, service pricing, and migration requirements all affect the comparison. Provider-specific interfaces or data formats can create vendor lock-in, making movement between environments costly or technically difficult. (nvlpubs.nist.gov)
Security and reliability
Cybersecurity in cloud environments involves a division of responsibilities. Providers generally protect facilities and underlying infrastructure; customers retain responsibilities for their data, permissions, and application configuration. The boundary varies by service. An IaaS customer may patch guest operating systems, whereas a managed database service transfers more maintenance work to the provider. (aws.amazon.com)
Data privacy and data governance involve access controls, data location, retention, and accountability. Technical controls include encryption grounded in cryptography, authentication, logging, and isolation between customers. Outsourcing infrastructure does not eliminate organizational responsibility for information handling. (aws.amazon.com)
Reliability also depends on network connectivity, provider operations, and application design. A service-level agreement specifies contractual service commitments, but does not make outages impossible. Backup, recovery, and redundant deployment remain distinct architectural concerns; the availability of cloud infrastructure alone does not establish an application’s recoverability. (nvlpubs.nist.gov)