Data centres and cloud computing connect the world by turning distant computing resources into services people can reach through networks. Search, messaging, streaming, online learning, business software and artificial intelligence may feel immediate on a screen, but behind that experience sit servers, storage systems, network links, power systems, cooling, software and people coordinating many handovers.
Did you know that “the cloud” is not a single place in the sky? Cloud computing is a way of organising computing resources so applications can use processing, storage and services across connected infrastructure. Understanding data centers, cloud computing, servers, networks and digital infrastructure becomes easier when we follow one request from a user device to a service and back again.
Find your next route: return to the Global Connectivity Hub to move between transport and logistics, digital networks, energy and industry, money and rules, science and health, education and knowledge, people and culture, or food, water and the environment.
Deepen the route through How the Internet Works, Semiconductors and Cybersecurity. Return to the eduKate Ecosystem Hub whenever the next question belongs to Mathematics, English, Science, evidence or learner repair. The numerical examples below are learning models rather than operational specifications for real facilities.
Begin with one request
Imagine a learner opening an online lesson. The visible action is a tap. The hidden journey may include a local wireless link, an internet provider, routing systems, a data centre, an application server, a database, storage, security checks and a return path. Draw only the nodes needed to explain the journey. Then label every arrow with a verb: requests, routes, authenticates, computes, stores, retrieves, returns.
This is the first big idea of global connectivity: the useful unit is not merely a machine. It is the relationship between machines, software, standards, networks and people. A powerful server that cannot be reached, powered, cooled, secured or understood is not a complete service.
What is a data centre?
A data centre is a facility designed to house computing equipment and the systems that keep it operating. The computing layer includes servers, storage and networking equipment. The support layer includes electricity, cooling, physical security, fire protection, monitoring and operational procedures. The building therefore matters, but the service depends on a much larger system than the building alone.
Servers perform work
A server is a computer that provides resources or services to other computers. It may run a website, process an application request, host a database, train a model or manage files. In large systems, workloads are distributed across many servers so one machine does not have to perform every job.
Storage preserves useful state
Some tasks are temporary; others need information to persist. Storage systems hold files, databases, backups and other data so a later request can continue from an earlier state. The important question is not only how much data can be stored, but how quickly, reliably and safely the correct data can be retrieved.
Networks move the request
Networks connect users, data centres and other services. A cloud application may depend on local fibre, long-distance terrestrial links, submarine cables and internet exchange points before a request reaches the chosen facility. This is why digital infrastructure is inseparable from the physical geography explained in the Internet room.
What cloud computing changes
Traditional computing often tied a task closely to a particular local machine. Cloud computing makes resources easier to allocate, combine and reach over networks. A team may rent computing capacity when needed instead of owning every server itself. Software can be deployed across multiple machines, regions or providers. The abstraction is useful because users can focus on the service, but good reasoning still asks what physical and organisational dependencies remain underneath.
Virtualisation separates logical from physical
One physical machine can support multiple logical computing environments, while one application can also span many physical machines. This separation helps resources be shared and adjusted. It also creates a learning trap: the interface can make infrastructure look effortless even though capacity, hardware, energy, maintenance and geography still matter.
Elasticity is controlled change
Cloud platforms can add or remove resources as demand changes. That does not mean capacity is infinite. Elasticity works because systems measure demand, allocate available resources and enforce limits. A good question is: which part can expand quickly, which part is fixed for longer, and what bottleneck appears first?
Latency changes the experience
Latency is the delay between an action and its response. Distance can contribute to delay, but so can routing, congestion, processing, storage access and software design. For an ordinary document, a small delay may be acceptable. For interactive communication, remote control or competitive gaming, the same delay can matter much more. Performance therefore depends on the job the connection must perform.
A Mathematics model of latency
Use an invented journey with four sequential delays: 18 milliseconds for access, 22 for routing, 35 for processing and 15 for the return stage. The total is 90 milliseconds. If processing falls from 35 to 20 while the other stages remain, the model becomes 75 milliseconds. The calculation is easy; the reasoning question is harder: was processing really the bottleneck, and are all stages truly sequential?
A Mathematics model of throughput
Suppose a fictional service must pass through stages that can handle 900, 650 and 1,100 requests per second. In a simple continuous-flow model, the completed flow cannot exceed 650 requests per second. Adding capacity to the 1,100 stage alone does not change that limit. This is the same bottleneck logic used across supply chains, transport and communications.
Capacity is more than processor speed
A system may be limited by processing power, memory, storage speed, network bandwidth, database design, power availability, cooling, software locks or human operations. When a service slows, “buy faster computers” is therefore not a diagnosis. Find the first constrained relationship before deciding what to change.
Redundancy and resilience are different
Redundancy means having alternatives or duplicates. Resilience means the wider system can continue or recover when something goes wrong. Two servers in one room provide some redundancy, but a common power failure may affect both. Two facilities in different places may reduce one shared risk while introducing coordination, cost and data-consistency challenges.
Backups need a return path
A backup is useful only if the right information can be restored when needed. That requires schedules, versions, permissions, integrity checks and tested recovery procedures. A folder named “backup” is not proof of recoverability. This principle transfers directly to student work: saved notes are valuable when the learner can retrieve and use them.
Power is part of digital connectivity
Servers require electricity, and dense computing equipment creates heat. That makes power supply, electrical distribution, cooling and facility design part of the digital system. A digital service can therefore depend on physical infrastructure that seems unrelated when we look only at the screen. Continue through Electricity Grids to make that dependency visible.
Cooling protects the operating range
Computing equipment converts electrical energy into useful work and heat. Cooling systems remove heat so equipment can operate within designed conditions. The engineering question is not simply “make it cold.” It is to control temperature, airflow, humidity, energy use, reliability and maintenance within an overall system.
Semiconductors sit beneath the cloud
Processors, memory, storage controllers and networking equipment depend on semiconductor supply networks. A cloud service can feel weightless, yet its physical capability begins with minerals, materials, fabrication, packaging, logistics and manufacturing. Follow Semiconductors and Critical Minerals to trace that deeper layer.
Standards make independent systems interoperable
Connected computing relies on agreed ways to represent, transmit, secure and interpret information. Standards do not make every system identical. They make important interfaces predictable enough for separately built components to cooperate. The Research and Inquiry Hub helps learners distinguish a standard, an implementation and an observed result.
Cybersecurity protects relationships
Security is not a decorative layer added after a service works. Identity, access control, encryption, logging, patching, network design and incident response determine who can use a system and what they are allowed to do. The purpose is not to create a system no one can reach; it is to make legitimate access possible while managing risk.
Data sovereignty and location questions
Organisations may care where data is stored, processed or backed up because legal, contractual, operational or risk requirements differ. The key learning move is to separate the physical location of equipment, the logical location of a service, the people responsible for it and the rules governing the information. Similar words can describe different layers.
Content delivery brings useful copies closer
Popular content can be copied or cached nearer to users so every request does not travel to the same distant origin. This can reduce delay and relieve pressure on central systems. But caching introduces a new question: how does the system know when a copy is stale? Connectivity often improves by adding copies, but copies also create a synchronisation problem.
Observability makes invisible systems inspectable
Operators use logs, metrics, traces and alerts to understand what a distributed system is doing. A single error message may be the visible end of a failure that began elsewhere. Good diagnosis follows the path of the request and asks which evidence belongs to each stage.
Failure often happens at interfaces
- a user can reach the network but not authenticate;
- an application can run but cannot reach its database;
- a database can respond but storage is saturated;
- a facility can operate but an external network route is unavailable;
- a backup can exist but permissions block restoration;
- a service can be technically available but too slow for the intended task.
Notice how none of these descriptions simply says “the cloud is down.” Precise language turns a vague failure into a testable mechanism.
More computing is not always the answer
If a queue forms because one database transaction is slow, adding many application servers may simply send more work toward the same constrained point. If demand is temporary, permanent overbuilding may also waste resources. Engineering improves when the response matches the actual constraint.
The human layer matters
Cloud systems are designed, maintained, documented and used by people. Skill differences affect whether a technically available platform becomes useful capability. Clear interfaces, training, accessibility, documentation and support are therefore part of connectivity. The SETC English Learning Library supports precise explanation, while the Well Being library helps keep access and belonging visible.
Access is not the same as capability
Two learners can be given the same online tool and obtain very different outcomes. One may understand file organisation, search, prompting, verification and privacy; another may not yet have those skills. Equal access to infrastructure does not automatically create equal realised capability. Education strengthens the human-to-technology interface.
Singapore as a connected learning specimen
A student in Singapore can use a familiar online service as a specimen without needing privileged technical access. Ask where the visible interaction begins, what information must travel, what could be cached locally, what probably requires remote computation and what evidence would be needed before making a claim. The goal is disciplined inference, not pretending to know the provider’s private architecture.
A paper cloud activity
Write five cards: user, network, application, database and storage. Pass a paper “request” through the cards. Add a timestamp, user identity and request ID. Then remove one piece of context and see what becomes ambiguous. Add a duplicate application card and ask whether the bottleneck moved. This simple exercise makes distributed systems visible without needing a server room.
Vocabulary should clarify relationships
- data centre — a facility housing computing and support infrastructure;
- cloud computing — network-accessible computing resources organised as services;
- server — a computer or process providing a service to others;
- latency — delay between an action and response;
- throughput — completed work per unit time;
- redundancy — additional components or paths that can provide alternatives;
- resilience — ability to continue, adapt or recover after disruption;
- cache — a stored copy used to make later access faster.
A student route
Start with one familiar application. Draw the request path, identify three possible bottlenecks, calculate one invented latency total and explain one resilience choice. Then close the notes and rebuild the explanation from memory. The Sengkang Learning Atlas supports diagnose, repair, retrieval and transfer when the explanation breaks.
A parent and teacher route
Ask the learner to explain the cloud without saying “it is just online.” Listen for physical infrastructure, network movement, computing, storage, limits and evidence. Reward precise relationships rather than a long list of technical terms. If a term is used without a mechanism, ask: what does it connect, and what changes when it fails?
Frequently asked questions
Is the cloud the same as the internet?
No. The internet provides interconnected networks that move data between systems. Cloud computing uses network connectivity to provide computing resources and services. The cloud depends on networks, but the two ideas are not identical.
Are data centres only about websites?
No. They can support storage, databases, business systems, scientific computing, communications, artificial intelligence, media delivery and many other workloads.
Does a closer data centre always mean a faster service?
Not always. Distance can affect latency, but routing, congestion, software, database behaviour, caching and device conditions also matter. Measure the whole path rather than assuming one variable explains everything.
Why can one failure affect many services?
Connected systems share dependencies. Several applications may use the same identity service, network route, cloud region or software component. Specialisation improves efficiency but can create common points of dependency.
Keep the return paths visible
Continue through How the Internet Works, Semiconductors, Electricity Grids, Cybersecurity and Standards and Measurement. Each room owns a different mechanism while the larger digital system stays visible.
A final connected-cloud investigation
Choose one everyday cloud service. Build a one-page explanation with one bounded diagram, one clearly labelled illustrative calculation, one likely physical dependency, one security dependency, one uncertainty and one question you would need evidence to answer. Give it to someone unfamiliar with cloud computing. Their first sensible question shows where your explanatory handover can improve.