Cloud Computing Basics Interview Questions with Answers
Last updated 2026-11-19
Cloud computing questions for freshers and early-career roles are conceptual: what the service models are, why the cloud scales, who is responsible for what, and how to keep a service available. Interviewers usually go beyond definitions and ask you to choose between options, such as when to use a virtual machine, a container or a serverless function, or to calculate downtime from an availability figure. This page answers those, with the availability and scaling arithmetic worked out.
The answers are provider-neutral; the vocabulary maps onto every major cloud. For the tools around it see Docker interview questions, AWS interview questions and Git interview questions, and for the role-level picture how to crack a DevOps interview.
| Service models | IaaS, PaaS, SaaS, plus FaaS (serverless) |
|---|---|
| Numericals | Downtime from availability, series and parallel availability, autoscaling target, monthly cost |
| Questions | 17 |
Who manages what
- On-premises: you manage everything from the building to the application.
- IaaS: the provider runs hardware, networking and the virtualization layer; you manage the operating system, runtime and application.
- PaaS: the provider also manages the operating system and runtime; you manage the application and data. SaaS: the provider manages everything except your configuration and data.
Cloud Computing Basics Interview Questions with Answers: questions with worked answers
1 What is cloud computing and what are its essential characteristics?
easyCloud computing delivers computing resources (servers, storage, databases, networking, software) over the internet on demand, paid for by use. The NIST definition lists five characteristics: on-demand self-service, broad network access, resource pooling (multi-tenant sharing of hardware), rapid elasticity (scale out and in quickly) and measured service (usage metered and billed). It turns capital expense into operating expense.
2 Explain IaaS, PaaS and SaaS with examples.
easyInfrastructure as a Service rents virtual machines, storage and networks and leaves the operating system and software to you, for example a virtual server instance. Platform as a Service provides a managed runtime where you deploy code without handling servers, for example an app hosting platform or a managed database. Software as a Service delivers a finished application through a browser, such as email, office suites and CRM systems. Function as a Service (serverless) runs individual functions on events and bills per execution.
3 What are the public, private, hybrid and multi-cloud deployment models?
easyA public cloud is owned by a provider and shared by many customers over the internet. A private cloud is dedicated to one organisation, on premises or hosted, for control and compliance. A hybrid cloud connects private and public environments so workloads can move between them, for example keeping sensitive data on premises and bursting compute to the cloud. Multi-cloud uses services from more than one provider to avoid lock-in or to use the best service of each.
4 What is the shared responsibility model?
mediumThe provider is responsible for security of the cloud (physical data centres, hardware, the hypervisor and managed service internals), and the customer is responsible for security in the cloud: identity and access management, data classification and encryption, network rules, operating system patching on IaaS, and application code. The split shifts with the service model, with the customer holding less under SaaS. Most breaches come from customer misconfiguration, such as public storage buckets and over-broad permissions.
5 What is virtualization and how do VMs differ from containers?
mediumVirtualization uses a hypervisor to run several virtual machines on one physical host, each with its own guest operating system and virtual hardware, giving strong isolation and the ability to run different operating systems. Containers share the host kernel and isolate processes using namespaces and cgroups, so they are far lighter, start in seconds or less and have little overhead, but the isolation is weaker and they must match the host kernel type. See Docker interview questions.
6 Differentiate scalability and elasticity, and vertical and horizontal scaling.
mediumScalability is the ability to handle more load by adding resources, while elasticity is doing it automatically and quickly both up and down as demand changes. Vertical scaling (scale up) makes one machine bigger with more CPU or memory: simple but limited by the largest machine and often needs downtime. Horizontal scaling (scale out) adds more instances behind a load balancer: nearly unlimited and fault tolerant, but the application must be stateless or share state externally.
7 Numerical: An autoscaling policy targets 50 percent average CPU. There are 4 instances running at 80 percent. How many instances are needed?
mediumDesired instances = ceil(current instances x current utilisation / target) = ceil(4 x 80 / 50) = ceil(6.4) = 7 instances. At 7 instances the same total load spreads to 4 x 80 / 7 = 45.7 percent each, which is below the target, whereas 6 instances would give 53 percent, above it. Real policies also use cool-down periods to avoid flapping and a minimum and maximum size.
8 Numerical: Convert availability of 99.9 percent, 99.99 percent and 99.999 percent into yearly downtime.
mediumA year has 365 x 24 x 60 = 525,600 minutes. 99.9 percent allows 0.1 percent downtime: 525.6 minutes, about 8.76 hours. 99.99 percent allows 0.01 percent: 52.56 minutes. 99.999 percent allows 0.001 percent: 5.26 minutes. Each extra nine cuts the allowed downtime by ten times and typically needs redundancy across independent failure domains, so the cost rises steeply.
9 Numerical: Two components in series each have 99.9 percent availability. What is the combined availability? What if two replicas with 99 percent availability each run in parallel?
mediumIn series both must work, so availability multiplies: 0.999 x 0.999 = 0.998001, about 99.8 percent, worse than either part (about 17.5 hours of downtime a year compared with 8.76). In parallel the system fails only when both fail: 1 - (0.01 x 0.01) = 0.9999 = 99.99 percent. This assumes independent failures, which is why replicas are placed in separate availability zones. A chain of many dependencies lowers overall availability, so redundancy must be added at every tier.
10 What are regions and availability zones?
mediumA region is a geographic area where a provider has data centres, such as Mumbai or Frankfurt. Each region has several availability zones, isolated locations with independent power, cooling and networking, connected by low-latency links. Deploying across zones protects against a data-centre failure, and across regions protects against a regional outage and also reduces latency for distant users, and may be needed for data residency rules.
11 Compare object, block and file storage.
mediumObject storage keeps data as objects with metadata in a flat namespace accessed over HTTP APIs: highly durable, very scalable and cheap, suited to backups, media and static websites, but not for low-latency in-place edits. Block storage presents raw volumes that a server formats with a file system, giving low latency for databases and boot disks. File storage provides a shared network file system (NFS or SMB) for many servers using the same files.
12 What is serverless computing and what are its trade-offs?
mediumIn serverless (function as a service) you deploy functions that the provider runs on demand and scales automatically, including to zero, and you pay per invocation and execution time, with no servers to manage. Trade-offs are cold-start latency, execution time and memory limits, statelessness, harder local testing, vendor-specific triggers and costs that can exceed always-on servers at steady high load. It suits event-driven and spiky workloads.
13 What is IAM and what is the principle of least privilege?
mediumIdentity and Access Management controls who (users, groups, roles, services) can do what (actions) on which resources, using policies. Least privilege means granting only the permissions needed for the task and nothing more, narrowing resources and actions, preferring roles with temporary credentials over long-lived access keys, enabling multi-factor authentication for people, and reviewing permissions regularly. Never use the root or administrator account for daily work.
14 Numerical: A virtual machine costs $0.10 per hour (an illustrative price). Find the monthly cost running continuously for 730 hours, and with a 40 percent discount for a one-year commitment.
mediumOn demand: 0.10 x 730 = $73 per month. With a 40 percent commitment discount the rate is 0.06 per hour: 0.06 x 730 = $43.80 per month, a saving of $29.20 (about $350 a year). If the machine is needed only 8 hours on 22 working days, on demand costs 0.10 x 176 = $17.60, cheaper than the committed instance, so commitments only pay off for steady workloads. Prices vary by provider and region.
15 What are RTO and RPO in disaster recovery?
mediumRecovery Time Objective is the maximum acceptable time to restore service after a disaster. Recovery Point Objective is the maximum acceptable amount of data loss, measured as time since the last recoverable copy. A one-hour RPO means backups or replication must run at least hourly. Strategies from cheapest and slowest to costliest and fastest are backup and restore, pilot light, warm standby and multi-site active-active.
16 What are load balancers, CDNs and auto scaling groups and how do they fit together?
hardA CDN serves cached static content from edge locations near users. A load balancer spreads incoming requests over healthy instances, checking health and removing failed ones, and can terminate TLS. An auto scaling group launches or terminates instances from a template based on metrics or schedules, registering them with the load balancer. Together they make a stateless web tier that scales out under load, tolerates instance failure and serves users with low latency.
17 How would you migrate an on-premises application to the cloud? Name the common strategies.
hardThe usual strategies are rehost (lift and shift the same VM images with little change), replatform (small changes such as moving to a managed database), refactor (redesign for cloud-native services such as containers and queues), repurchase (replace with SaaS), retire and retain. Steps: assess dependencies and data volumes, choose the strategy per workload, set up networking, identity and landing-zone security, migrate data with minimal downtime using replication, test, cut over with a rollback plan, then optimise cost. See how to crack a DevOps interview.
Frequently asked questions
Which cloud provider should a fresher learn first?
Pick one major provider and learn the shared concepts: compute, storage, networking, IAM, monitoring. The concepts map across providers, and a single hands-on project on any of them is worth more than shallow knowledge of all.
What is the difference between cloud computing and virtualization?
Virtualization is a technology that divides physical hardware into virtual machines. Cloud computing is a service model built on virtualization (and automation, metering and self-service) that delivers resources on demand over a network.
What is vendor lock-in?
Dependence on one provider's proprietary services that makes moving costly. It is reduced by using open standards, containers, portable data formats and abstraction layers, at the cost of not using every managed feature.
What does pay-as-you-go mean?
Billing is based on measured use, such as instance hours, storage per GB-month and requests, with no upfront hardware purchase. Costs follow usage but can surprise you without budgets and alerts.
Related guides
- Docker Interview Questions and Answers
- AWS Interview Questions and Answers
- DevOps Interview Questions and Answers
- Git Interview Questions and Answers with Commands
- Linux Interview Questions and Answers: Concepts, Permissions and Processes
- Computer Networks Interview Questions with Worked Answers and Numericals
- Operating Systems Interview Questions with Worked Answers
More in CS Fundamentals Interview Questions
- Computer Architecture Interview Questions with Numericals
- Computer Networks Interview Questions with Worked Answers and Numericals
- CPU Scheduling Interview Questions with Gantt Chart Numericals
- Database Indexing Interview Questions
- DBMS Concurrency Control Interview Questions
- DBMS Interview Questions and Answers
- DBMS Keys Interview Questions
- Deadlock Interview Questions with Banker's Algorithm Worked Example
- Design Patterns Interview Questions with Code
- Distributed Systems Interview Questions
- DNS Interview Questions: Resolution, Record Types, Caching and Troubleshooting
- ER Model Interview Questions