{
  "id": "infrastructure-as-a-service",
  "title": "Infrastructure as a Service",
  "lang": "en",
  "source_lang": "en",
  "summary": "Infrastructure as a Service is a cloud computing model that supplies on-demand computing, storage, and networking resources while customers manage operating systems and applications.",
  "category": "computer-science",
  "subcategory": "software",
  "type": "concept",
  "aliases": [
    "IaaS",
    "Infrastructure-as-a-Service"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/infrastructure-as-a-service/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/infrastructure-as-a-service/",
      "json": "https://aiwiki.page/zh/infrastructure-as-a-service.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Infrastructure as a Service (IaaS) is a cloud computing service model in which a provider supplies processing capacity, storage, networking, and related computing resources that customers provision and manage on demand. Customers can deploy operating systems and applications without owning or directly maintaining the underlying physical infrastructure. The provider manages the physical equipment and its supporting facilities, while customers generally control their operating systems, software, data, and selected network configurations. Resources are commonly metered and billed according to consumption, allocated capacity, or contractual commitments."
    },
    {
      "heading": "Defining characteristics",
      "text": "IaaS provides infrastructure as a programmable service rather than as equipment that an organization must purchase, install, and maintain. Customers typically provision resources through a web console, a command-line interface, or an application programming interface (API).\n\nIts defining characteristics follow those of cloud computing:\n\n- On-demand self-service: customers can provision resources without a provider employee manually fulfilling each request.\n- Network access: services are available through network connections, including public or private connections.\n- Resource pooling: infrastructure serves multiple consumers, with resources assigned and reassigned as demand changes.\n- Rapid elasticity: capacity can expand or contract relatively quickly, subject to quotas, hardware availability, and service limits.\n- Measured service: usage is monitored and accounted for, supporting consumption-based charging and capacity management.\n\nRenting a remotely hosted server does not, by itself, establish that a service is IaaS. Traditional hosting may lack automated provisioning, elastic capacity, or pooled resource management. Conversely, IaaS does not require every resource to be virtualized: providers may offer dedicated physical servers through the same automated provisioning and metering systems."
    },
    {
      "heading": "Core resources and architecture",
      "text": "Compute\n\nCompute services provide processing power and memory. A common unit is the virtual machine, often called an instance, created from an image containing an operating system and, optionally, preinstalled software.\n\nVirtualization enables multiple isolated virtual machines to share a physical server. A virtualization layer allocates processor time, memory, and access to storage and network devices. Customers usually choose among instance configurations optimized for general-purpose workloads, memory capacity, processor performance, or specialized acceleration.\n\nSome offerings include graphics processing units for scientific computing, graphics rendering, and machine learning. Others provide bare-metal servers when workloads require direct hardware access, particular licensing arrangements, or stronger separation from other customers.\n\nStorage\n\nIaaS commonly includes several storage models:\n\n- Block storage exposes volumes that behave like disks and can hold operating systems, file systems, and application data.\n- Object storage stores data as objects accessed through service interfaces, often for media, archives, logs, and backups.\n- File storage provides shared access using file-oriented protocols and directory structures.\n- Local or ephemeral storage is associated with a particular host or instance and may not survive its replacement or termination.\n\nThese models differ in access methods, performance, persistence, and pricing. Storage persistence and redundancy should not be confused with backup: redundant copies can protect against some hardware failures without protecting against accidental deletion or unwanted changes.\n\nNetworking\n\nNetworking services connect resources to each other, to customer premises, and to external networks. Typical features include logically isolated networks, subnets, routing, public and private addresses, traffic filtering, and private connectivity.\n\nLoad balancing distributes requests across multiple servers, while network controls restrict which systems can communicate. Customers configure these logical facilities without normally administering the provider’s underlying physical switches and routers.\n\nControl and resource management\n\nThe provider’s control plane handles resource creation, configuration, identity checks, quotas, and accounting. Workloads run in a corresponding execution environment, often described as the data plane. A control-plane disruption can prevent management operations even when some existing workloads continue running.\n\nThe physical resources are housed in data centers. Providers often organize infrastructure into geographic regions and smaller deployment areas intended to separate certain failures. The meaning and degree of independence of these areas vary by provider."
    },
    {
      "heading": "Responsibility and service boundaries",
      "text": "IaaS follows a shared responsibility model, although the precise division depends on the service and contract.\n\nArea | Typical provider responsibility | Typical customer responsibility\nFacilities and hardware | Buildings, power, cooling, physical security, servers, and physical networking | Choosing suitable services and deployment locations\nVirtualization | Maintaining and securing the virtualization platform | Configuring virtual resources appropriately\nOperating systems | Infrastructure needed to run customer instances | Guest operating-system configuration, updates, and administration\nApplications | Availability of contracted infrastructure services | Application deployment, maintenance, and security\nData and identities | Provider-side protection mechanisms and service controls | Data classification, permissions, credentials, and use of security features\n\nThis arrangement gives customers substantial control but leaves them responsible for significant operational work. An unpatched application or an overly permissive network rule generally remains a customer-side issue even when the provider’s infrastructure operates correctly.\n\nThe boundary is also service-specific. A virtual machine and a managed database offered by the same provider do not transfer the same responsibilities to that provider."
    },
    {
      "heading": "Comparison with other service models",
      "text": "In Platform as a Service (PaaS), the provider manages more of the software environment, such as application runtimes and deployment platforms. Customers concentrate more directly on application code and data rather than on operating-system administration.\n\nIn Software as a Service (SaaS), the customer uses a complete application operated by the provider. Control is generally exercised through application settings and supported integrations rather than through underlying infrastructure.\n\nServerless computing further abstracts server provisioning for supported execution environments. Servers still exist, but customers usually interact with functions, jobs, or managed services rather than persistent server instances.\n\nThese categories describe responsibility boundaries, not necessarily separate vendors or physical systems. A single application may combine IaaS virtual machines, a managed database, serverless functions, and SaaS components."
    },
    {
      "heading": "Provisioning, elasticity, and automation",
      "text": "IaaS enables infrastructure to be managed through software. An automated deployment can create networks, storage volumes, virtual machines, and traffic-routing rules as a coordinated process.\n\nInfrastructure as code represents intended infrastructure in machine-readable definitions. Keeping those definitions in version control supports review, repeatability, and reconstruction. It does not eliminate configuration drift or deployment errors, but makes infrastructure changes easier to track and automate.\n\nElasticity commonly takes two forms:\n\n- Vertical scaling: changing the capacity of an individual resource, such as assigning more memory or processors.\n- Horizontal scaling: changing the number of resources, such as adding application servers.\n\nAutomatic scaling uses metrics, schedules, or policies to adjust capacity. Effective scaling still depends on application design. A workload with a single bottleneck, persistent local state, or strict coordination requirements may not benefit simply from adding machines. Techniques from distributed computing are often needed to use elastic resources effectively.\n\nElasticity is not unlimited or instantaneous. Resource shortages, startup times, service quotas, and regional restrictions can constrain provisioning."
    },
    {
      "heading": "Deployment and pricing",
      "text": "IaaS can be delivered through public, private, or hybrid cloud environments. Public-cloud providers offer services to many customers; private clouds provide cloud capabilities for the exclusive use of one organization. Hybrid deployments connect distinct environments while retaining mechanisms for moving or integrating workloads and data.\n\nPublic IaaS commonly uses several pricing mechanisms:\n\n- On-demand pricing, based on allocated resources or measured consumption.\n- Commitment-based discounts, exchanged for spending or usage commitments.\n- Interruptible capacity, offered at lower prices with the possibility of reclamation by the provider.\n- Dedicated capacity, charged for exclusive access to designated hardware or resource pools.\n\nCharges may apply separately to compute, storage capacity, storage operations, network traffic, public addresses, and supporting services. An idle allocated server may continue to incur charges, and stopping it may leave chargeable storage or network resources in place.\n\nIaaS can shift expenditure from purchasing equipment toward ongoing service consumption. It is not inherently cheaper than owning infrastructure: the outcome depends on utilization, staffing, workload variability, discounts, migration costs, and data-transfer charges."
    },
    {
      "heading": "Uses, reliability, and limitations",
      "text": "Common uses include website hosting, enterprise applications, development and testing, batch processing, scientific computation, disaster recovery, and temporary capacity expansion. IaaS is particularly relevant when customers need control over operating systems or must run software that cannot easily move to a more managed platform.\n\nReliability depends on both the service and the customer’s architecture. A service-level agreement defines contractual commitments and remedies; it is not a guarantee that an application will never fail. Infrastructure redundancy, application failover, backups, and recovery procedures address different failure modes.\n\nSecurity depends partly on effective access control, authentication, and encryption. Automation can reduce inconsistent manual configurations, but compromised credentials or erroneous scripts can also propagate changes rapidly across many resources.\n\nOther limitations include:\n\n- Performance variability: shared infrastructure and network conditions can affect latency and throughput.\n- Network dependence: remote resources depend on connectivity, and geographic distance imposes latency.\n- Operational complexity: customers must manage systems, monitor workloads, investigate failures, and control spending.\n- Data-location requirements: replication, backups, and service configuration can affect data privacy and jurisdictional obligations.\n- Vendor lock-in: proprietary interfaces, network designs, identity systems, and data-transfer costs can complicate migration."
    },
    {
      "heading": "Historical development",
      "text": "IaaS developed from earlier practices in time-sharing, utility computing, remote hosting, and virtualization. Its distinctive contribution was to combine remotely supplied infrastructure with self-service interfaces, automated provisioning, elastic resource allocation, and metered consumption.\n\nAmazon’s launch of Elastic Compute Cloud in 2006 was an important milestone in commercial public IaaS. Other providers subsequently expanded the market, while private-cloud software brought comparable provisioning models into organizational data centers. In 2011, the United States National Institute of Standards and Technology formalized IaaS as one of three principal cloud service models, alongside PaaS and SaaS.\n\nThe model has since expanded beyond general-purpose virtual machines to include accelerated computing, dedicated hardware, and increasingly programmable storage and networking."
    }
  ],
  "links": [
    "cloud-computing",
    "operating-system",
    "application-programming-interface",
    "virtual-machine",
    "virtualization",
    "graphics-processing-unit",
    "machine-learning",
    "data-backup",
    "load-balancing",
    "data-center",
    "platform-as-a-service",
    "software-as-a-service",
    "serverless-computing",
    "infrastructure-as-code",
    "version-control",
    "distributed-computing",
    "service-level-agreement",
    "access-control",
    "authentication",
    "encryption",
    "data-privacy",
    "vendor-lock-in"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}