Question 1

What is cloud computing?

Accepted Answer

Cloud computing is the delivery of computing resources — servers, storage, databases, networking, software — over the internet, on demand, paid for by usage rather than bought up front. Instead of buying and running hardware, you rent capacity from a provider and scale it up or down as needed. A good fresher answer names the shift from capital expenditure to operating expenditure, and from weeks to provision a server to seconds. Adding one concrete example — "we can create a virtual machine in a minute instead of ordering hardware" — makes it real rather than recited.

Question 2

What are the main benefits of using the cloud?

Accepted Answer

Elasticity — scale up for a traffic spike and back down after; pay for what you use rather than for peak capacity; global reach without building data centres; managed services that remove operational work; faster time to market; and built-in reliability through multiple zones and regions. Interviewers also want honesty about the trade-offs: less control, potential vendor lock-in, and the fact that cloud can be more expensive than on-premises for a steady, fully-utilised workload. A candidate who names one downside sounds far more credible than one who lists only benefits.

Question 3

What are the different cloud deployment models?

Accepted Answer

Public cloud, where infrastructure is shared and provided by a vendor such as Google, AWS or Azure; private cloud, dedicated to one organisation either on premises or hosted; hybrid, combining private infrastructure with public cloud; and multi-cloud, using more than one public provider. The reason to know these is that interviewers often follow with "why would a bank use hybrid?" — the answer being regulation, data residency and existing hardware investment. Having one concrete reason ready for each model is what makes the answer useful.

Question 4

What is the difference between IaaS, PaaS and SaaS?

Accepted Answer

IaaS gives you raw infrastructure and you manage the operating system upwards — Compute Engine. PaaS gives you a managed runtime and you supply only code and configuration — App Engine or Cloud Run. SaaS is finished software you just use — Gmail or Google Workspace. The higher up you go, the less you manage. Giving a GCP example for each is what turns a textbook answer into a GCP answer. A useful closing line is that the trade-off is control versus convenience: more control means more responsibility for patching, scaling and availability.

Question 5

What is Google Cloud Platform?

Accepted Answer

Google Cloud Platform is Google's public cloud, offering compute, storage, networking, databases, big data, machine learning and management services on demand, billed by usage. It runs on the same global infrastructure that powers Google Search, YouTube and Gmail. It is worth naming two or three flagship products — Compute Engine, BigQuery, GKE — because a candidate who can only say "it is Google's cloud" has clearly not used it. Mentioning that Kubernetes originated at Google is a nice touch.

Question 6

What is a region and a zone in GCP?

Accepted Answer

A region is a geographic area such as asia-south1 in Mumbai. A zone is an isolated location within a region, effectively a separate failure domain, and each region has three or more. You deploy across multiple zones to survive a zone failure and across regions to survive a region failure. The follow-up is always about high availability, so pair the definition with the reason: spreading virtual machines across zones means a single data centre problem does not take your application down. That cause-and-effect is what interviewers are listening for.

Question 7

What is a GCP project?

Accepted Answer

A project is the container for all your resources and the boundary for billing, quotas, APIs and access control. Every resource belongs to exactly one project. Each project has a globally unique project ID that cannot be changed, a display name, and a Google-assigned project number. A useful detail is that teams typically use separate projects for development, staging and production, because the project is the cleanest isolation boundary GCP offers. Knowing the project ID is permanent is the small fact that shows you have created one.

Question 8

What is Compute Engine?

Accepted Answer

Compute Engine is GCP's virtual machine service. You choose a machine type, an operating system image, disks and network settings, and you get a VM you control from the operating system upwards, billed per second after a one-minute minimum. Two GCP-specific facts worth adding: per-second billing with automatic sustained-use discounts, and live migration, where Google moves a running VM to another host during maintenance with no reboot. Both are genuine differentiators.

Question 9

What is Cloud Storage?

Accepted Answer

Cloud Storage is GCP's object storage service. You store files as objects inside globally-named buckets and access them over HTTP. It is used for backups, static website assets, data lakes, images and logs, and it scales without you provisioning capacity. The point to make is that it is object storage, not a filesystem — there are no real folders, and the slashes in an object name are just part of the key. Candidates who describe it as a mounted disk usually get caught on the next question.

Question 10

What are the Cloud Storage classes?

Accepted Answer

Standard for frequently accessed data; Nearline for data accessed about monthly, with a 30-day minimum; Coldline for roughly quarterly access, with a 90-day minimum; and Archive for data accessed less than once a year, with a 365-day minimum. Storage cost falls at each step while retrieval cost rises. The detail that impresses is the minimum storage duration: deleting an Archive object after a week still bills 365 days. That is why lifecycle rules should move data down the tiers rather than a script deleting and rewriting it.

Question 11

What is a VPC?

Accepted Answer

A Virtual Private Cloud is your private network inside GCP, where your resources get internal IP addresses and communicate securely. In GCP the VPC is global — it spans every region — and subnets within it are regional. The global VPC is GCP's most distinctive networking feature and a very common interview question, because in AWS a VPC is regional. Being able to state that difference confidently is worth a lot in a GCP interview.

Question 12

What is IAM in GCP?

Accepted Answer

Identity and Access Management controls who can do what on which resource. You grant a role — a bundle of permissions — to a principal such as a user, group or service account, on a resource. Permissions are never granted directly, always through roles. Adding that roles come in three kinds — basic, predefined and custom — and that basic roles like Owner and Editor are too broad for production, shows you understand the practice and not just the mechanism.

Question 13

What is a service account?

Accepted Answer

A service account is an identity for an application, virtual machine or workload rather than a person. It is identified by an email address and is used so code can call Google Cloud APIs with its own permissions, without a human logging in. The best-practice point to add is that you should attach a service account to the resource rather than downloading a JSON key file, because a downloaded key is a long-lived credential that can leak. Volunteering that unprompted is a strong signal.

Question 14

What is BigQuery?

Accepted Answer

BigQuery is GCP's serverless data warehouse. You load data in and query it with standard SQL, and it can scan terabytes in seconds. There are no servers or clusters to manage, and you pay for the data your query scans plus the data you store. The cost fact worth knowing is that SELECT * costs far more than selecting specific columns, because BigQuery is columnar and only reads the columns you reference. That is the most-asked BigQuery detail at every level.

Question 15

What is Google Kubernetes Engine?

Accepted Answer

GKE is GCP's managed Kubernetes service. Google runs and secures the Kubernetes control plane and, in Autopilot mode, the nodes as well, so you deploy containerised applications and Kubernetes handles scheduling, scaling, self-healing and rolling updates. Mentioning that Kubernetes originated at Google, from an internal system called Borg, is a nice piece of context. Being able to name the difference between Autopilot and Standard mode shows you have looked beyond the definition.

Question 16

What is a container and how is it different from a virtual machine?

Accepted Answer

A container packages an application with its dependencies and shares the host operating system kernel, so it is lightweight and starts in seconds. A virtual machine includes a full guest operating system on top of a hypervisor, so it is heavier, slower to start, but more isolated. The practical consequence to state is density and speed: you can run many more containers than VMs on the same hardware, and start them far faster, which is what makes container-based autoscaling responsive.

Question 17

What is Cloud Run?

Accepted Answer

Cloud Run runs stateless containers without you managing any servers. You give it a container image that listens on a port, and it handles scaling from zero to many instances, HTTPS, and billing per request-second — so an idle service costs nothing. The phrase to use is scale to zero, because that is what makes it different from a VM or a Kubernetes node that costs money whether or not it serves traffic. Being able to say when a container is the wrong fit — stateful workloads — shows balance.

Question 18

What is Cloud SQL?

Accepted Answer

Cloud SQL is GCP's managed relational database service for MySQL, PostgreSQL and SQL Server. Google handles provisioning, patching, backups, replication and failover, while you use normal SQL and standard client drivers. The distinction interviewers often check is Cloud SQL versus BigQuery: Cloud SQL is for transactional application data with many small reads and writes, BigQuery is for analytical queries over large volumes. Mixing them up is a common fresher mistake.

Question 19

What is the difference between horizontal and vertical scaling?

Accepted Answer

Vertical scaling makes one machine bigger — more CPU or memory. Horizontal scaling adds more machines behind a load balancer. Cloud platforms favour horizontal scaling because it gives both elasticity and fault tolerance, whereas a single bigger machine is still a single point of failure. The requirement to mention is that horizontal scaling needs the application to be stateless, or to keep its state somewhere shared. That is why cloud-native applications avoid storing session data on the local machine.

Question 20

What is a load balancer?

Accepted Answer

A load balancer distributes incoming traffic across several backend servers so no single one is overwhelmed, and it stops sending traffic to a backend that fails a health check. On GCP the global external Application Load Balancer gives you a single IP address that serves users worldwide. The GCP-specific point is the single global anycast IP: users are routed to the nearest healthy backend automatically, without DNS tricks. That is genuinely different from having one load balancer per region.

Question 21

What is autoscaling?

Accepted Answer

Autoscaling automatically adds or removes instances based on a signal such as CPU utilisation, request rate or a custom metric, within a minimum and maximum you configure. It keeps performance acceptable during spikes and reduces cost when demand falls. A good addition is the cool-down period: without one, the autoscaler reacts to the CPU spike of a machine still booting and adds more instances than needed. Knowing that shows you have thought about how it behaves rather than just what it does.

Question 22

What is high availability?

Accepted Answer

High availability is designing a system so it keeps working when a component fails — by removing single points of failure, running redundant instances across zones, using health checks and automatic failover. It is measured as a percentage of uptime, such as 99.9% or 99.99%. A useful concrete example on GCP is a regional managed instance group spread across three zones behind a load balancer. Being able to describe one real configuration beats reciting the definition.

Question 23

What is the difference between availability and durability?

Accepted Answer

Availability is whether you can access the data or service right now. Durability is whether the data still exists and is intact over time. Cloud Storage, for example, offers extremely high durability — data is replicated so it is essentially never lost — separately from its availability SLA. The example that makes it click is that a storage service could be temporarily unreachable, so availability is affected, while the data itself remains perfectly safe. Freshers who conflate the two usually cannot answer follow-ups about backups.

Question 24

What is Cloud Shell?

Accepted Answer

Cloud Shell is a free, browser-based Linux environment with the gcloud command-line tool, kubectl, Terraform, Python and an editor already installed, authenticated as your account. It gives you 5 GB of persistent home directory storage and is the quickest way to run commands without installing anything. The detail to include is that only the home directory persists — anything installed elsewhere is lost when the session VM is recycled. That is a small practical fact you only know if you have used it.

Question 25

What is the gcloud command-line tool?

Accepted Answer

gcloud is the primary command-line interface for creating and managing GCP resources and for authentication. Related commands include gcloud storage for Cloud Storage and bq for BigQuery. Anything you can do in the Cloud Console you can generally do with gcloud. Being able to write one real command from memory — creating a VM or listing buckets — is worth more than describing the tool. Interviewers frequently ask for a command as a quick check of hands-on experience.

Question 26

What is an API and why must you enable APIs in GCP?

Accepted Answer

An API is the programmatic interface to a service. In GCP every service is fronted by an API that must be explicitly enabled on a project before use, because quotas, billing and audit logging are tracked per project and disabled APIs cannot be called at all. Recognising the error message "API not enabled on project" immediately is the practical value of knowing this, because it is one of the first errors anyone hits when starting a new project.

Question 27

What is a firewall rule in GCP?

Accepted Answer

A firewall rule controls which network traffic is allowed to or from your virtual machines. Each rule has a direction, a priority, an action to allow or deny, source or destination ranges, protocols and ports, and a target such as all instances or instances with a specific network tag. The two defaults to know are that all incoming traffic is denied and all outgoing traffic is allowed unless you change it. Knowing GCP firewall rules are stateful — return traffic is automatically allowed — is the other fact interviewers check.

Question 28

What is the difference between an internal and an external IP address?

Accepted Answer

An internal IP is a private address used for communication inside your VPC and is not reachable from the internet. An external IP is a public address that makes a resource reachable from the internet. External IPs can be ephemeral, changing when the instance stops, or reserved as static. The security practice to mention is that production VMs usually have no external IP at all, reaching the internet through Cloud NAT for outbound traffic and receiving traffic only through a load balancer. That is a strong answer for a fresher.

Question 29

What is a snapshot in Compute Engine?

Accepted Answer

A snapshot is a backup of a persistent disk, stored incrementally so only changed blocks are saved after the first one. You can restore a disk from a snapshot or use it to create new disks, and snapshots are stored globally so they can be restored in another region. The incremental behaviour is the useful detail: subsequent snapshots are cheap, and deleting an older snapshot does not break newer ones because Compute Engine reorganises the data automatically.

Question 30

What happens to your data when you delete a VM?

Accepted Answer

The boot disk is deleted with the instance by default, because auto-delete is enabled on it. Additional attached disks default to auto-delete off and survive. Local SSD data is always lost. Snapshots taken earlier are stored separately and are unaffected. The differing default between boot disks and data disks is exactly what makes this a good question. Knowing you can also enable deletion protection on an instance to prevent accidental removal rounds it out.

Question 31

What is Cloud Monitoring and Cloud Logging?

Accepted Answer

Cloud Monitoring collects numeric metrics such as CPU, latency and request count, and drives dashboards and alerts. Cloud Logging collects log entries, lets you search them, and can route them to other destinations. They are separate services used together to understand what a system is doing. The simple distinction to state is that metrics are numbers over time and logs are text records of events. Adding that you alert on metrics and investigate with logs shows you understand how they are used together.

Question 32

What is Pub/Sub?

Accepted Answer

Pub/Sub is a managed messaging service. Publishers send messages to a topic and subscribers receive them through subscriptions, so the sender and receiver are decoupled — the publisher does not need to know who consumes the message or whether they are currently running. The practical example that lands is an order-placed event triggering payment, inventory and email services independently. Knowing that delivery is at-least-once, so consumers must handle duplicates, is the detail that impresses at fresher level.

Question 33

What is serverless computing?

Accepted Answer

Serverless means you deploy code or a container and the platform handles provisioning, scaling and capacity entirely — you never size or patch a server and you pay only while your code runs. Cloud Run, Cloud Functions and BigQuery are GCP examples. The honest caveat to add is cold starts: when there are no warm instances, the first request waits for one to start. Naming a trade-off rather than presenting serverless as free of downsides is what makes the answer credible.

Question 34

What is the difference between Cloud Functions and Cloud Run?

Accepted Answer

Cloud Functions runs a single function in a supported language, deployed from source with no container to build. Cloud Run runs any container you provide, so you control the runtime and can serve multiple endpoints. Cloud Functions second generation is actually built on Cloud Run. The practical guidance is functions for small event handlers and Cloud Run for real services, and knowing that the two have converged technically is a current, accurate detail that many candidates miss.

Question 35

What is the difference between a bucket and a folder in Cloud Storage?

Accepted Answer

A bucket is a real container with a globally unique name, a location and access settings. Folders are not real — object names simply contain slashes, and the console displays them as a folder structure for convenience. Everything inside a bucket is a flat set of objects. The practical consequence is that renaming a "folder" means rewriting every object under that prefix, because there is nothing to rename. That surprises people coming from a filesystem background and is a good detail to know.

Question 36

How is GCP billed?

Accepted Answer

By usage — compute per second, storage per GB per month, network egress per GB, and per-operation or per-request charges for many services. A billing account is linked to projects, and you can set budgets with alerts and export detailed cost data to BigQuery. The fact worth adding is that budgets only alert, they do not stop spending. Knowing that, and knowing ingress is free while egress is charged, covers the two billing facts freshers are most often asked.

Question 37

What is the Google Cloud free tier?

Accepted Answer

It has two parts: a 90-day trial credit for new customers, and the Always Free tier giving perpetual monthly allowances on around 20 products, such as one small VM in specific US regions, 5 GB of Cloud Storage and 1 TB of BigQuery query processing per month. The catch worth knowing is that Always Free allowances are region-restricted for several products, so a free-tier VM must be in a supported US region. Anyone who has actually used the free tier usually learned this from an unexpected charge.

Question 38

What is a managed service and why would you use one?

Accepted Answer

A managed service is one where the provider handles the infrastructure, patching, scaling and availability — Cloud SQL, BigQuery, Cloud Run. You use one to reduce operational work and human error, at the cost of less control and some version or feature restrictions. Naming the trade-off honestly is what makes this a good answer: managed services cost more per unit but remove work, and the right choice depends on whether your team's scarcest resource is money or engineering time.

Question 39

What is the shared responsibility model?

Accepted Answer

Google is responsible for the security of the cloud — physical data centres, hardware, the hypervisor and the network. The customer is responsible for security in the cloud — access control, firewall rules, data classification, application code and, where an operating system is exposed, patching it. The point that shows understanding is that the boundary moves depending on the service: with Compute Engine you patch the operating system, with Cloud Run you do not. Naming that variation is what turns a memorised diagram into an actual answer.

Question 40

How would you deploy a simple web application on GCP? Walk me through it.

Accepted Answer

Containerise the application, push the image to Artifact Registry, and deploy it to Cloud Run, which gives an HTTPS endpoint that scales automatically. Store data in Cloud SQL or Firestore, keep secrets in Secret Manager, put a global load balancer with Cloud CDN in front for a custom domain, and use Cloud Monitoring and Logging for observability. This is the standard closing question for a fresher screen and it is testing whether you can connect several services into a working system. Naming where the data, the secrets and the monitoring go — not just the compute — is what makes it a complete answer.

GCP Interview Questions for Freshers Interview Questions and Answers

How to use this set