Under the Hood of the Cloud: How Global Cloud Infrastructure Works
How does Mr. Cloud Service Provider 🥸 distribute and organize his global infrastructure to serve millions of users worldwide?
Although cloud services appear virtual, they ultimately depend on physical equipment deployed and operated in data centres, including compute servers, storage devices, cooling systems, and more.
This article explains how Cloud Service Providers (CSPs) organize their infrastructure into geographic regions and, typically, multiple Availability Zones within a region. Without this architecture, delivering reliable cloud services to users across different parts of the world would be much more challenging. For example, an entertainment service like Netflix might be available in only a few countries, leaving users in other parts of the world out of the fun. 🤷♂️
Cloud Global Infrastructure
Cloud global infrastructure is the worldwide network of data centres, servers, storage systems, and communication networks that enables cloud providers to deliver computing services globally.
Its primary goals include global accessibility, scalability, high availability, reliability, and efficient service delivery. Together, these capabilities enable users and businesses to access applications, store data, and process information from almost anywhere in the world without owning physical servers.
Simple Real-Life Example
Imagine a company based in Canada that has customers in Canada, India, and the United Kingdom.
Without global cloud infrastructure, the company might need to build and maintain its own servers in several locations.
With cloud global infrastructure, the company can:
- Host its application in selected cloud regions.
- Deliver content closer to its customers.
- Replicate data across locations where appropriate.
- Scale computing resources as customer demand increases.
- Improve resilience by deploying applications across multiple Availability Zones.
This helps the company serve customers internationally without having to build its own worldwide network of data centres.
Main Components of Cloud Global Infrastructure
Cloud global infrastructure consists of the following major components**.**
Data Centres
Data centres are physical facilities that contain servers, storage devices, networking equipment, and power and cooling systems.
Functions:
- Store large amounts of data.
- Run applications and websites.
- Process and manage information.
- Provide computing resources to users.
Region
A region is a specific geographical area where a cloud provider operates its cloud infrastructure.
Examples include cloud regions in:
- Canada
- The United States
- The United Kingdom
- India
- Japan
A single country may have multiple cloud regions, depending on the provider.
Cloud providers allow customers to select a region based on their performance requirements, data residency needs, and business objectives.
Availability Zones (AZs)
An Availability Zone is an isolated location within a cloud region, typically consisting of one or more data centres.
Availability Zones are designed to provide a degree of isolation from failures affecting other zones in the same region.
Benefits:
- Improve application availability.
- Reduce the impact of individual infrastructure failures.
- Support disaster recovery within a region.
For example, an application can run across multiple Availability Zones so that it can continue operating if one zone experiences an outage.
However, deploying an application across multiple zones does not automatically guarantee high availability. The application must be configured to handle failures, and its supporting components must also be designed appropriately.
Edge Locations
Edge locations are infrastructure points positioned closer to end users to help deliver content and services with lower latency.
Benefits:
- Faster content delivery.
- Reduced response times.
- Improved user experience.
For example, a video streaming service can cache video content closer to viewers to reduce loading times.
This means a viewer may receive cached video content from a nearby edge location instead of having to retrieve it from the application's original hosting infrastructure every time.
Points of Presence (PoPs)
A Point of Presence (PoP) is a location where a cloud provider's network connects with users, internet service providers, or other networks.
PoPs can:
- Improve network connectivity.
- Help route traffic efficiently.
- Support content delivery and security services.
Although PoPs and edge locations may overlap in practice, the terms are not interchangeable in every cloud provider's architecture. A PoP describes a network presence, while an edge location often refers to infrastructure used to bring computing or content-delivery services closer to users.
Global Network
A global network connects data centres, cloud regions, and edge locations through high-speed communication links.
It enables:
- Data transfer between locations.
- Communication between cloud services.
- Global access to applications.
- Reliable connectivity for businesses and users.
How the Pieces Work Together
Now that we've explored the individual components, let's see how they work together in a real-world scenario.
Imagine a user in India opening a video-streaming application hosted on a cloud platform.
Here's a simplified explanation of what might happen:
- The user sends a request: The user opens the application, and their request travels through their internet connection toward the service.
- The network directs the request: Internet routing and the service's network configuration help direct traffic toward the appropriate destination.
- An edge location may deliver the content: If the requested video is cached at a nearby edge location, it may be delivered directly from there, reducing latency.
- The application processes the request: If the request requires application processing, it reaches the application's infrastructure in a cloud region.
- The application accesses supporting services: The application may communicate with databases, storage systems, or other services to retrieve the required information.
- The infrastructure supports reliability: Depending on how the application is designed, load balancers, redundant components, and deployments across multiple Availability Zones can help it handle traffic and withstand certain failures.
The exact process depends on the application's architecture and the cloud services being used.
The important thing to understand is that these components work together to deliver cloud services to users around the world. Each plays a different role in making applications accessible, responsive, and reliable.
Benefits of Cloud Global Infrastructure
- Global accessibility: Users can access cloud services from different parts of the world.
- Scalability: Computing resources can be increased or decreased according to demand.
- High availability: Redundant infrastructure helps keep applications running.
- Disaster recovery: Data and applications can be replicated across locations when configured appropriately.
- Low latency: Deploying resources closer to users can reduce response times.
- Cost efficiency: Businesses can reduce the need to purchase and maintain their own infrastructure.
- Security: Cloud providers offer security tools and controls to help protect data and applications.
Application Infrastructure: Scalability, Availability, and Disaster Recovery
An enterprise's cloud infrastructure must do more than simply run its applications. It must also support continuous operations and reliable service delivery to its customers.
A well-designed cloud infrastructure should be able to handle growth, maintain service availability, and recover from failures.
Three key capabilities help achieve these objectives:
- High Availability
- Scalability
- Disaster Recovery and Fault Tolerance
It's important to understand that cloud providers supply the underlying infrastructure and services, but customers must often configure their applications to take full advantage of these capabilities.
For example, a provider may offer multiple Availability Zones, but an application deployed in only one zone may still become unavailable if that zone experiences an outage.
Let's explore these capabilities in more detail.
High Availability
High availability ensures that applications and services remain accessible to users with minimal downtime, even when individual components fail.
Typical techniques used to achieve high availability include:
- Deploying application instances across multiple Availability Zones: Cloud Service Providers generally recommend multi-AZ architectures for applications with high availability requirements. This helps reduce disruptions caused by the failure of an individual Availability Zone.
- Distributing traffic across healthy instances: Load balancers can direct incoming requests to healthy application instances.
- Replicating databases where appropriate: Database replication can help improve availability and support recovery when failures occur.
- Avoiding single points of failure: Designing systems with redundancy helps prevent the failure of a single component from bringing down the entire application.
Real-life scenario: What happens if an Availability Zone fails?
Imagine an online shopping application running in two Availability Zones within the same cloud region.
Under normal conditions, both zones help handle customer requests.
Now, suppose one zone experiences a power or network failure.
If the application has been configured appropriately, the load balancer can stop directing traffic to unhealthy instances and send requests to healthy instances in the surviving zone.
The application may continue operating while the affected zone recovers.
However, this depends on whether the surviving infrastructure has sufficient capacity and whether the application's databases, storage, and other dependencies can continue functioning.
This illustrates why deploying across multiple Availability Zones can improve availability, although it does not eliminate every possible cause of downtime.
Scalability
Scalability is the ability of a system to handle changes in workload by adjusting its resources. This may involve adding or removing resources or increasing or decreasing their capacity.
Scaling can be categorized in two ways.
Based on the direction of resource changes
- Scaling Up: Increasing the resources or capacity available to handle a higher workload.
- Scaling Down: Decreasing resources or capacity when the workload reduces.
Based on how resources are changed
- Vertical Scaling (Scaling Up or Down): Increasing or decreasing the capacity of an existing resource, such as adding more RAM or CPU power to a server or cloud instance.
- Horizontal Scaling (Scaling Out or In): Adding more servers or instances to handle increased demand, or removing them when they are no longer needed.
For example, a website experiencing a surge in visitors might add more application instances behind a load balancer to handle the increased traffic.
This allows the website to serve more users without relying entirely on a single server with greater capacity.
Disaster Recovery and Fault Tolerance
Disaster recovery and fault tolerance help cloud infrastructure withstand failures, outages, and disasters while minimizing data loss and service interruptions.
Although these concepts are related, they are not identical.
Fault tolerance is the ability of a system to continue operating, potentially without interruption, even when certain components fail.
Disaster recovery focuses on restoring applications, services, and data following a major disruption.
Two key metrics used to measure the disaster recovery capabilities of the cloud infrastructure supporting an application or service are:
- Recovery Time Objective (RTO) — RTO is the maximum acceptable time an application or service can remain unavailable following a disruption before it must be restored.
- Recovery Point Objective (RPO) — RPO is the maximum acceptable amount of data loss, measured in time. It indicates how far back data recovery may need to go following a disruption.
In simple terms:
- RTO measures how quickly the system must recover.
- RPO measures how much data loss is acceptable.
For example, if an application has an RTO of two hours, it should be restored within two hours following a disruption.
If its RPO is 15 minutes, the recovery strategy should aim to limit data loss to changes made within the 15 minutes preceding the disruption.
Achieving these objectives depends on factors such as backups, data replication, recovery procedures, and the application's overall architecture.
Challenges of Expanding CSP Global Infrastructure
A Cloud Service Provider (CSP) may be prevented from establishing a component of its cloud infrastructure in a particular region due to several technical, economic, legal, environmental, and political constraints.
Imagine a cloud provider identifying a country with millions of potential customers.
Expanding into that market might seem like an obvious business opportunity. However, establishing reliable cloud infrastructure requires more than finding a suitable location and installing servers.
The provider must evaluate electricity availability, network connectivity, regulatory requirements, physical security, environmental conditions, and expected customer demand.
Even when a region is commercially attractive, these challenges can make deployment impractical or significantly more expensive.
Major Challenges and Limitations
High Infrastructure Costs
Establishing data centres requires substantial investment in land, buildings, servers, networking equipment, power systems, cooling systems, and security.
If the expected demand for cloud services in a region is too low, the investment may not be financially viable.
Insufficient Power Supply
Data centres require a reliable and continuous supply of electricity.
Regions with limited power-generation capacity, unreliable electricity grids, or expensive electricity may be unsuitable for large-scale cloud infrastructure.
Poor Network Connectivity
A region may lack the high-speed internet connections, fibre-optic networks, and reliable communication infrastructure required to support cloud services.
This can increase latency and reduce service performance.
Legal and Regulatory Restrictions
Governments may impose restrictions concerning data protection, data residency, privacy, cybersecurity, and foreign ownership.
A CSP may be unable to operate in a region if it cannot satisfy local laws or obtain the necessary licences and approvals.
Political Instability and Security Risks
Political conflicts, civil unrest, sanctions, and geopolitical tensions can threaten infrastructure, disrupt operations, or make investment too risky.
Physical security threats may also discourage deployment.
Environmental and Climatic Conditions
Extreme temperatures, floods, earthquakes, hurricanes, and other natural hazards can increase construction costs and the risk of service disruptions.
Water scarcity and environmental regulations may also make it difficult to provide adequate cooling for data centres.
Limited Availability of Skilled Personnel
Cloud infrastructure requires qualified professionals to install, maintain, secure, and operate hardware and software.
A shortage of skilled workers in a region can increase operating costs and make reliable service delivery more difficult.
Insufficient Market Demand
If a region has few potential customers or limited demand for cloud services, the CSP may not generate enough revenue to justify the initial investment and ongoing operational expenses.
Lack of Suitable Land and Infrastructure
Suitable sites may be unavailable because of land scarcity, high property prices, zoning restrictions, inadequate transport infrastructure, or difficulties obtaining construction permits.
Difficulty Maintaining Reliability and Redundancy
A CSP must ensure that its infrastructure can withstand equipment failures, network disruptions, and power outages.
If a region lacks suitable facilities for backup power, redundant connectivity, or disaster recovery, maintaining the required level of service availability may be difficult.
Final Verdict
The cloud is not an absence of infrastructure. It is physical infrastructure organized, abstracted, networked, and managed to deliver computing resources on demand.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article