What is Cloud Platforms?

Cloud platforms are online computing environments that let you provision processing power, storage, databases, networking and managed application services without owning the underlying data-centre hardware. Instead of buying a server for every workload, you request resources from a provider and reach them over a network. The important distinction is that a cloud platform is not simply remote storage: it is a programmable pool of computing services that can be created, scaled, measured and released on demand.

For UK broadband users, the cloud also changes where work happens. Your laptop, server or home lab may be local, but the application, data or compute can sit in a provider region elsewhere. That means download speed, upload speed, bandwidth and latency can all affect the experience in different ways.

What makes a platform a cloud platform?

NIST defines cloud computing around five essential characteristics: on-demand self-service, broad network access, resource pooling, rapid elasticity and measured service. Those characteristics matter because they separate genuine cloud behaviour from simply renting a fixed server in somebody else's building.

• On-demand self-service: resources can be provisioned when needed without a manual provider process.

• Broad network access: services are available over networks using standard mechanisms.

• Resource pooling: provider resources serve multiple consumers while maintaining logical separation.

• Rapid elasticity: capacity can expand or contract quickly as demand changes.

• Measured service: usage can be monitored and metered.

Primary source: NIST SP 800-145, The NIST Definition of Cloud Computing.

What does a cloud platform provide?

• Compute: virtual machines, containers, accelerators and other processing resources.

• Storage: object, block and file storage for applications, backups and datasets.

• Databases: managed relational, NoSQL, cache and analytics services.

• Networking: virtual networks, routing, firewalls, load balancers, private connectivity and public endpoints.

• Managed services: monitoring, identity, messaging, queues, security tools, AI services and application runtimes.

The attraction is not that every cloud service is automatically faster or cheaper. It is that infrastructure and higher-level services can be consumed programmatically, scaled without installing physical hardware and combined into architectures that would be difficult to reproduce on a single local machine.

IaaS, PaaS, SaaS and serverless: what is the difference?

• Infrastructure as a Service (IaaS): you provision fundamental resources such as compute, storage and networking. You usually manage more of the operating system, applications and configuration.

• Platform as a Service (PaaS): the provider manages more of the runtime and underlying infrastructure while you focus on your application and data.

• Software as a Service (SaaS): you consume a finished application rather than building on the underlying platform.

• Serverless or Function as a Service: you deploy code or functions while the platform handles server provisioning and much of the scaling. It is commonly treated as a cloud-platform component rather than a replacement for the broader NIST service-model definitions.

The UK National Cyber Security Centre makes a useful practical distinction: SaaS is normally a pre-built service that solves a business problem, while cloud platforms provide components used to build services, including IaaS, PaaS and serverless components.

See the NCSC introduction to cloud security for its current terminology and security guidance.

How does a cloud platform work?

A request normally starts on a user device or another server. DNS and routing direct traffic across local networking, the broadband or enterprise connection, transit and peering networks, and finally into the provider's infrastructure. Inside the cloud, virtual networking and load balancing steer the request to the appropriate compute or managed service. Storage or database operations may happen behind that application before a response returns.

1. User device or local server

2. Router, firewall and local Ethernet or WiFi

3. UK broadband or business internet connection

4. Internet routing and provider edge

5. Cloud region and virtual network

6. Compute, storage, database or managed service

7. Return path to the user or another system

The slowest or most constrained relevant part of that path can become the practical bottleneck. A fast cloud instance cannot compensate for a saturated upload, poor WiFi or a high-latency path to the chosen region.

Cloud regions and availability zones

Cloud providers divide infrastructure geographically. A region is a broad geographic location containing cloud infrastructure. Availability zones are separate locations within a region designed to provide fault isolation. The exact implementation and service availability differ by provider.

Current UK examples show why the detail matters. AWS announced a fourth Availability Zone for its Europe (London) eu-west-2 Region in August 2026. Microsoft currently lists UK South in London with three availability zones and UK West in Cardiff without availability-zone support. These are provider-specific capabilities and can change, so architecture should be checked against current provider documentation.

AWS London Region announcement and Microsoft Azure region list.

Public cloud, private cloud, hybrid cloud and multi-cloud

• Public cloud: services are offered from provider infrastructure to many customers with logical separation.

• Private cloud: cloud infrastructure is dedicated to one organisation, whether on premises or hosted elsewhere.

• Hybrid cloud: distinct private and public cloud environments are connected so data or applications can move or interact between them.

• Multi-cloud: an organisation deliberately uses services from more than one cloud provider. Multi-cloud is a common architectural term, but it is not one of the four deployment models in the original NIST SP 800-145 definition.

Hybrid and multi-cloud can improve flexibility or reduce dependence on one environment, but they also increase networking, identity, monitoring, data-transfer and operational complexity.

Why broadband matters to cloud platforms

• Download speed affects how quickly large datasets, backups, software images and other data can be received from the cloud.

• Upload speed affects backups, media uploads, synchronisation and any workload that sends substantial data to the cloud.

• Latency affects interactive applications and workflows involving many request-and-response cycles, even when little data is transferred.

• Packet loss and unstable WiFi can trigger retransmissions, stalls and inconsistent application performance.

• Router, firewall, Ethernet, WiFi and device limits can constrain the connection before traffic reaches the ISP.

This is why a cloud application can feel slow even when the cloud provider itself is healthy. End-to-end performance depends on the complete path, not one headline broadband speed.

Techie Broadband calculation: uploading 10 GB to the cloud

Theoretical example using decimal units: 10 GB equals 80 gigabits. Transfer time in seconds is 80,000 megabits divided by the available upload rate in megabits per second.

• 20 Mbps upload: 80,000 ÷ 20 = 4,000 seconds, or 66 minutes 40 seconds.

• 100 Mbps upload: 80,000 ÷ 100 = 800 seconds, or 13 minutes 20 seconds.

• 500 Mbps upload: 80,000 ÷ 500 = 160 seconds, or 2 minutes 40 seconds.

• 1 Gbps upload: 80 gigabits ÷ 1 gigabit per second = 80 seconds.

• 2.5 Gbps upload: 80 ÷ 2.5 = 32 seconds.

These are theoretical line-rate examples, not real-world promises. Protocol overhead, encryption, application behaviour, storage speed, provider throttling, competing traffic and the actual broadband product all increase transfer time. Many UK broadband products are asymmetric, so the upload rate may be much lower than the advertised download rate.

Techie Broadband calculation: why latency can matter more than bandwidth

Consider a workflow that requires 100 sequential network round trips before it can complete. At a 20 ms round-trip time, the network component alone contributes about 2 seconds. At 80 ms, it contributes about 8 seconds. The data volume could be tiny in both cases.

This simplified calculation assumes each operation waits for the previous round trip. Modern applications often parallelise requests, cache data or use persistent connections, so real applications will differ. The point is that adding bandwidth does not automatically remove delay caused by sequential network interactions.

Cloud platforms vs self hosting

A cloud platform moves infrastructure responsibility towards a provider. Self hosted services run on hardware you control, such as a home server or NAS. Neither approach is universally better.

• Cloud platform: rapid provisioning, elastic capacity and a large catalogue of managed services, but ongoing usage charges and provider dependency.

• Self hosting: direct hardware and data control, but you own more responsibility for power, connectivity, updates, backups and physical failures.

• Hybrid approach: keep some services local while using cloud storage, off-site backup, managed databases or public endpoints where they add value.

For local infrastructure concepts, see What is Home Lab?, What is Docker? and What is Virtual Machine?.

Cloud security and the shared responsibility model

Using a cloud provider does not outsource every security decision. The provider and customer share responsibility, and the split changes with the service model. With IaaS, the customer normally controls and secures more of the operating system and workload. With a more managed service, the provider takes on more of the underlying stack, but the customer still has responsibilities for areas such as identity, access, data and configuration.

• Use strong identity controls and multi-factor authentication where available.

• Apply least-privilege access rather than broad administrator permissions.

• Know which services are public and which are private.

• Encrypt sensitive data in transit and at rest where appropriate.

• Log and monitor important administrative and application events.

• Back up data independently where recovery requirements demand it.

• Review provider guidance because secure defaults and available controls change.

The NCSC cloud security principles provide a UK-focused framework for evaluating cloud services and the responsibilities that remain with the customer.

Does CGNAT or a dynamic IP stop you using cloud platforms?

Usually no. Ordinary cloud consumption is outbound: your device opens a connection to a public cloud endpoint, so CGNAT or a changing residential public IPv4 address normally does not prevent access. The situation changes when the cloud needs to initiate a connection back to equipment on your premises or when you are building site-to-site networking.

In those cases, technologies such as outbound tunnels, VPNs, IPv6 or provider-specific private connectivity can be relevant. IPv6 can provide globally routable addressing, but firewall policy and application security remain essential.

How to troubleshoot a slow cloud platform

1. Confirm whether the problem affects one cloud service or all internet activity.

2. Test a wired Ethernet connection to remove WiFi as a variable.

3. Check download and upload throughput separately.

4. Measure latency and packet loss to a relevant destination where testing is permitted.

5. Pause large backups, synchronisation jobs and other competing traffic.

6. Check router WAN and LAN interface rates and CPU load if available.

7. Test another device to separate local device limitations from network behaviour.

8. Check the cloud provider's service-health information for the relevant region and service.

9. Check application logs, database performance and storage limits before blaming the network.

10. Compare another time or network path if the issue is intermittent.

A cloud performance problem can sit in the application, provider service, storage layer, region, internet path, ISP, router, WiFi or end device. Troubleshooting is more reliable when each layer is isolated in order.

When should you use a cloud platform?

• When workloads need to scale without purchasing hardware first.

• When managed databases, queues, monitoring or serverless components reduce operational work.

• When services need regional deployment or resilient architecture.

• When teams need infrastructure that can be automated through APIs and infrastructure as code.

• When off-site compute or storage is useful for backup, development, analytics or public applications.

Cloud is not automatically the right answer for every workload. Predictable local services, privacy-sensitive systems, very large data transfers or applications that must continue during an internet outage may justify local or hybrid designs.

Common misconceptions about cloud platforms

• The cloud is not 'nowhere': workloads run on physical infrastructure in data centres.

• Cloud does not mean unlimited speed: network, service and account limits still exist.

• Cloud does not remove security responsibility: configuration and identity remain critical.

• A faster broadband package does not automatically reduce latency to a cloud region.

• Using a UK cloud region does not guarantee every dependency or replicated copy stays in the same location; service-specific data residency and architecture must be checked.

Cloud Platforms FAQs

Are AWS, Microsoft Azure and Google Cloud cloud platforms?

Yes. They are major public cloud platforms offering compute, storage, networking, databases and many managed services. Their product names, regions, pricing and capabilities differ.

Is cloud storage the same as a cloud platform?

No. Cloud storage is one category of cloud service. A cloud platform can also provide compute, networking, databases, identity, monitoring, application runtimes and many other components.

Do cloud platforms need fast broadband?

They need reliable connectivity, but the required speed depends on the workload. Large backups and datasets benefit from high upload and download throughput. Interactive services can be more sensitive to latency and packet loss than raw bandwidth.

Can I use cloud platforms behind CGNAT?

Yes for most normal outbound cloud access. CGNAT becomes more relevant when external systems need to initiate inbound connections to equipment on your local network.

Is the cloud more secure than self hosting?

Not automatically. Large providers can offer extensive security capabilities, but the result depends on the service, architecture and customer configuration. The NCSC recommends understanding the shared responsibility model and evaluating the provider against appropriate security requirements.

Key takeaway

A cloud platform is a programmable set of remote computing resources and managed services that you can provision over a network. It can remove the need to own every piece of physical infrastructure, but it does not remove networking, security or architecture from the equation. For a UK user, the practical experience still depends on the broadband path, upload and download capacity, latency, local equipment and the cloud region and service you choose.

To explore the related technical concepts used in this Cloud Platforms guide, visit our broadband terminology guide.