What are the key components of cloud architecture?
Cloud computing architecture is the set of parts that together deliver computing as an on-demand service over a network. There is no single standard list of components. Textbooks and vendors group them differently, so the most reliable starting point is the NIST definition (NIST SP 800-145, 2011). It describes three things:
- Five essential characteristics: on-demand self-service, broad network access, resource pooling, rapid elasticity, and measured service.
- Three service models: Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS). Function as a Service (serverless) is usually treated as a later extension.
- Four deployment models: public, private, community, and hybrid. Multi-cloud, meaning the use of more than one provider, is not one of the NIST deployment models.
Those definitions describe what the cloud offers. The architecture describes how the pieces are assembled to deliver it. A common way to look at it is in layers, and the sections below follow that structure.
The front end
The front end is what users and client systems touch. It includes client devices such as laptops and phones, web browsers, apps, and the interfaces used to request services: the web console, the command-line interface (CLI), software development kits (SDKs), and APIs.
The network
The internet, or a private connection, links the front end to the back end and carries data between users and cloud resources. Network design affects latency, security, and availability. Load balancers and virtual networks sit in this layer. See how load balancers distribute traffic and how a virtual private cloud isolates a network.
The back end
The back end is the provider’s side of the system. It contains:
- Physical servers, storage, and networking equipment in data centers.
- A virtualization layer, which uses a hypervisor to divide physical hardware into virtual machines. Virtualization makes cloud computing practical, but it is not the same thing as cloud computing.
- A platform and application layer, which runs services such as databases, messaging, and runtime environments.
- Management software for orchestration, monitoring, metering and billing, autoscaling, and identity and access management (IAM).
- Security controls that protect the infrastructure and the services running on it.
A common layered view runs from hardware, to virtualization, to platform and middleware, to application.
Regions and availability zones
Cloud providers organize their physical infrastructure into regions. A region contains availability zones (AZs). An AZ is one or more data centers with separate power and networking. Spreading an application across AZs protects it from the failure of a single data center. Protection against the failure of a whole region requires a second region. Disaster recovery explains how these choices fit into a recovery plan.
Compute, storage, and services
Compute can come from virtual machines, containers, or serverless functions. Containers and virtual machines differ in isolation and startup time, and providers offer many types of virtual machine for different workloads.
Storage usually comes in three types:
| Type | What it stores | Examples |
|---|---|---|
| Block | Raw volumes attached to a virtual machine or server | AWS EBS, Azure Managed Disks |
| Object | Files and data addressed by key and accessed over HTTP | Amazon S3, Azure Blob Storage, Google Cloud Storage |
| File | Shared file systems that many clients can mount | Amazon EFS, Azure Files, Google Cloud Filestore |
Management and automation
Management services handle the work that keeps a cloud environment running and billable:
- Orchestration and provisioning create and arrange resources. Provisioning and task scheduling are covered in a separate article.
- Monitoring and logging track health and performance.
- Metering and billing record usage so that customers pay for what they use.
- Autoscaling adds or removes capacity as demand changes.
- Identity and access management controls who can do what.
Security and shared responsibility
Security is split between the provider and the customer. The provider secures the underlying infrastructure. The customer secures its own data, identities, and configuration. The customer’s share is largest in IaaS and smallest in SaaS, but it does not disappear in SaaS: data, access, and identity remain the customer’s responsibility.
Service agreements describe what the provider commits to. How an SLA is set up, measured, and enforced is a separate topic.
Design principles
AWS’s Well-Architected Framework is one widely used checklist. It has six pillars: operational excellence, security, reliability, performance efficiency, cost optimization, and sustainability. Sustainability was added in 2021. The framework belongs to one provider, but its questions apply to any cloud design.
Standard roles in the NIST reference architecture
The NIST cloud computing reference architecture (SP 500-292, 2011) names five roles: cloud consumer, cloud provider, cloud carrier, cloud auditor, and cloud broker. These roles help explain who is responsible for what in a contract or an audit.
Putting the pieces together
Cloud computing consists of front-end access, a network, and a back end of virtualized, pooled infrastructure and services, all managed through software that provisions, meters, and secures them. Each provider names these parts differently, so comparing architectures works best by function (compute, storage, networking, identity, and management) rather than by product name.