Cloud Computing: On-Demand Infrastructure, Platforms, and Services
Cloud computing delivers computing resources — servers, storage, databases, networking, software — over the internet on a pay-as-you-go basis. Instead of owning physical hardware, organizations rent capacity from providers like AWS, Azure, and Google Cloud. For telecom, cloud is the foundation of modern network functions: 5G cores run as microservices in containers, and SDN/NFV replace dedicated hardware with software on commodity servers.
Service Models — IaaS, PaaS, SaaS
Cloud services are grouped into three layers. Each layer abstracts more of the underlying infrastructure from the user:
The provider supplies raw compute (virtual machines), storage (block, object, file), and networking (virtual switches, load balancers). You manage the OS, runtime, middleware, and applications. Examples: AWS EC2, Azure Virtual Machines, Google Compute Engine. Best for teams that need full control over the stack — e.g., running a custom Linux kernel or specialized database engine.
The provider manages OS, runtime, patching, and scaling. You deploy your application code and data. Examples: AWS Elastic Beanstalk, Google App Engine, Azure App Service, Heroku. Best for developers who want to focus on writing code without worrying about server management or OS updates.
The provider manages everything — infrastructure, platform, and the application itself. You use the software through a browser or API with no installation or maintenance. Examples: Gmail, Microsoft 365, Salesforce, Zoom. The most widely used cloud model for end users; the provider handles upgrades, backups, and availability.
| You Manage | IaaS | PaaS | SaaS |
|---|---|---|---|
| Application | ✓ | ✓ | ✗ |
| Data | ✓ | ✓ | ✓ (yours) |
| Runtime / Middleware | ✓ | ✗ | ✗ |
| Operating System | ✓ | ✗ | ✗ |
| Servers / Storage / Network | ✗ | ✗ | ✗ |
Deployment Models
Resources are owned and operated by a third-party provider and shared across many customers (multi-tenant). Lowest upfront cost, virtually unlimited scale, and global reach. Security and compliance responsibility is shared between provider and customer.
Resources are dedicated to a single organization, either on-premises or hosted by a provider. Full control over hardware, data sovereignty, and compliance. Higher capital cost but required for sensitive workloads (e.g., core banking, defence).
Combines private and public clouds connected by secure links (VPN or dedicated circuits like AWS Direct Connect). Sensitive workloads stay on-premises; burstable or less-sensitive workloads run in public cloud. Most enterprises operate hybrid today.
Using services from two or more public cloud providers. Avoids vendor lock-in, lets teams pick best-of-breed services, and improves resilience. Increases operational complexity — requires unified tooling (Terraform, Kubernetes) to manage consistently.
Core Concepts
A hypervisor (e.g., VMware ESXi, KVM, Hyper-V) runs multiple virtual machines (VMs) on a single physical server. Each VM has its own OS and appears to software as a dedicated machine. Virtualization is the foundational technology that made cloud computing economically viable — one server can host tens or hundreds of VMs.
Containers (e.g., Docker) package an application and its dependencies into a lightweight, portable unit that shares the host OS kernel. Faster to start than VMs (milliseconds vs. seconds), use less memory, and are ideal for microservices. Kubernetes orchestrates containers at scale — scheduling, load balancing, auto-scaling, and self-healing across clusters of nodes.
Functions as a Service (e.g., AWS Lambda, Azure Functions, Google Cloud Functions) lets you run code without provisioning servers. You upload a function; the platform executes it in response to events (HTTP request, file upload, queue message) and charges per invocation. Scales to zero when idle — no charges when not running. Best for event-driven, short-lived tasks.
Unlike block storage (disk) or file storage (NFS), object storage (e.g., AWS S3, Azure Blob Storage) stores data as flat objects with metadata and a unique key. Infinitely scalable, highly durable (11 9s on S3), and cheap. The backbone for storing backups, media files, logs, ML training datasets, and static website assets.
Cloud in Telecom — NFV and SDN
Telecom networks historically ran on proprietary, purpose-built hardware. Cloud computing changed this through two key transformations:
NFV replaces dedicated network appliances (firewalls, routers, IMS nodes, EPC components) with software running as virtual network functions (VNFs) on standard x86 servers. A single server rack can host VNFs that previously required racks of proprietary hardware. The ETSI NFV framework defines the NFVI (infrastructure), VNF layer, and MANO (management and orchestration). In 5G, all core network functions (AMF, SMF, UPF) are cloud-native network functions (CNFs) running in containers.
SDN separates the control plane (decisions about where traffic goes) from the data plane (forwarding packets). A centralized SDN controller (e.g., OpenDaylight, ONOS) programs forwarding rules into commodity switches via OpenFlow or NETCONF/YANG. This enables dynamic, automated traffic engineering — slicing bandwidth for different tenants, re-routing around failures in milliseconds — capabilities that are essential for 5G network slicing.
- 5G Core (5GC) is fully cloud-native — all NFs (AMF, SMF, UPF, PCF, UDM) are microservices communicating over HTTP/2 (SBI), deployable on Kubernetes
- MEC — Multi-access Edge Computing — runs cloud workloads at the base station or aggregation site, reducing latency to <5 ms for real-time applications (AR/VR, autonomous vehicles)
- O-RAN — Open RAN disaggregates the radio access network into open interfaces, allowing cloud-based RAN management (Near-RT RIC, Non-RT RIC) to optimize radio resources with ML
- Network Slicing — SDN + NFV together allow a single physical 5G network to be partitioned into multiple virtual networks (slices), each with its own QoS, latency, and security profile
Key Cloud Providers
Launched in 2006, AWS is the largest cloud provider. Key services: EC2 (compute), S3 (storage), RDS (managed databases), Lambda (serverless), VPC (networking), CloudFront (CDN). Offers the broadest service catalogue and the most mature global infrastructure (33 regions).
Azure integrates tightly with Microsoft products (Active Directory, Office 365, Teams). Strong in hybrid scenarios via Azure Arc. Key services: Azure VMs, Azure Kubernetes Service (AKS), Azure DevOps, Cosmos DB. Preferred by enterprises already running Microsoft stacks.
Google invented Kubernetes (open-sourced 2014) and leads in data analytics (BigQuery) and AI/ML (Vertex AI, TPUs). Key services: GKE (managed Kubernetes), Cloud Spanner (globally distributed DB), Pub/Sub (messaging), Cloud Run (serverless containers). Strong choice for data-intensive and ML workloads.
OpenStack is an open-source cloud operating system used to build private and telco clouds. Core components: Nova (compute), Neutron (networking), Swift (object storage), Cinder (block storage). Widely deployed by telecom operators as the NFVI platform for hosting VNFs.
Key Concepts at a Glance
- Elasticity — automatically scale resources up during peak demand and down during idle periods; pay only for what you use
- High Availability (HA) — cloud providers spread resources across multiple Availability Zones (AZs) within a region; a single datacenter failure does not affect the application
- Regions and Availability Zones — a region is a geographic area (e.g., eu-west-1); an AZ is one or more datacenters within that region with independent power and networking
- CDN — Content Delivery Network — distributes static content (images, videos, JS bundles) to edge servers close to users, reducing latency globally (e.g., AWS CloudFront, Cloudflare)
- IAM — Identity and Access Management — fine-grained control over who (users, services, roles) can perform which actions on which resources; fundamental to cloud security
- CapEx vs. OpEx — cloud shifts spending from capital expenditure (buying hardware) to operational expenditure (monthly bills), which improves cash flow and enables faster iteration
Abbreviations
| Abbreviation | Full Form | Abbreviation | Full Form |
|---|---|---|---|
| IaaS | Infrastructure as a Service | PaaS | Platform as a Service |
| SaaS | Software as a Service | FaaS | Function as a Service |
| VM | Virtual Machine | VNF | Virtual Network Function |
| NFV | Network Functions Virtualization | NFVI | NFV Infrastructure |
| SDN | Software Defined Networking | MANO | Management and Orchestration |
| CNF | Cloud-Native Network Function | MEC | Multi-access Edge Computing |
| AZ | Availability Zone | CDN | Content Delivery Network |
| IAM | Identity and Access Management | VPC | Virtual Private Cloud |
| K8s | Kubernetes | SBI | Service-Based Interface (5G) |
| O-RAN | Open Radio Access Network | RIC | RAN Intelligent Controller |
Why Cloud Matters for Telecom
- 5G Core is cloud-native by design — 3GPP mandated a service-based architecture for 5GC; all NFs are containerized microservices from the standard itself
- Operational cost reduction — replacing purpose-built appliances with VNFs on commodity servers cuts hardware costs by 60–70% and reduces power consumption
- Faster service launch — new network features deploy as software updates in minutes, not hardware upgrades taking months
- Edge computing enables new use cases — MEC brings cloud compute to within one hop of the user, making ultra-low-latency applications (real-time gaming, robotic surgery, V2X) viable
- AI/ML integration — cloud platforms provide the compute (GPUs, TPUs) and data pipelines needed to train and run ML models that optimize radio networks, predict faults, and detect fraud in real time