Junior Backend Architecture Interview Questions (50 MCQs)

Question 1

In a layered architecture (controller/service/repository), what is the main responsibility of the controller?

Accepted Answer

Receiving the HTTP request, validating input, and calling the service

Answer

Running SQL queries directly against the database

Answer

Holding all business rules and domain calculations

Answer

Managing the connection pool and database transactions

Question 2

What is the purpose of the Repository pattern?

Accepted Answer

Hiding data-access details behind a collection-like interface

Answer

Caching HTTP responses to reduce server load

Answer

Managing different API versions for clients

Answer

Distributing requests across multiple service instances

Question 3

What does the Separation of Concerns principle mean?

Accepted Answer

Each part of the system handles one distinct concern

Answer

Every service must run on a separate physical server

Answer

Frontend and backend code must be written in different languages

Answer

Every team must own its own dedicated database

Question 4

What is Dependency Injection?

Accepted Answer

Dependencies are supplied to a class from outside instead of it creating them itself

Answer

Classes build their dependencies with `new` inside their own constructor

Answer

Injecting code at runtime to patch bugs without deploying

Answer

Automatically installing npm packages at build time

Question 5

What does the Dependency Inversion Principle (the D in SOLID) state?

Accepted Answer

High- and low-level modules should both depend on abstractions

Answer

High-level modules should depend directly on concrete implementations

Answer

Each class should have only one reason to change

Answer

Subclasses must be substitutable for their base class

Question 6

What is a monolith in backend architecture?

Accepted Answer

An application deployed as a single unit with all features in one process

Answer

A set of small services communicating through messages

Answer

An application made only of serverless functions

Answer

A system with a separate database for each user

Question 7

What is a key benefit of microservices compared to a monolith?

Accepted Answer

Independent deployment and scaling of each service by separate teams

Answer

Automatically lower latency because all calls go over the network

Answer

Simpler debugging and monitoring of the whole system

Answer

Completely removing the need for transactions and consistency handling

Question 8

Which is a common cost of moving to microservices?

Accepted Answer

More operational complexity: networking, monitoring, and distributed data

Answer

Being unable to use more than one programming language

Answer

Inability to scale services horizontally

Answer

Having to deploy all services together on every change

Question 9

What is the difference between horizontal and vertical scaling?

Accepted Answer

Horizontal adds more machines; vertical makes one machine more powerful

Answer

Horizontal adds CPU and RAM; vertical adds servers

Answer

Horizontal is only for databases and vertical only for app servers

Answer

They are the same and only named differently in the cloud

Question 10

Why does being stateless matter for horizontal scaling?

Accepted Answer

Any instance can serve any request because no state lives in local memory

Answer

Because stateless services don't need a database

Answer

Because stateless services are always faster than stateful ones

Answer

Because being stateless removes the need for a load balancer

Question 11

What is the main job of a load balancer?

Accepted Answer

Distributing incoming traffic across multiple healthy instances

Answer

Permanently storing user data

Answer

Compiling code and building container images

Answer

Encrypting data at rest in the database

Question 12

How does the Round Robin load-balancing algorithm work?

Accepted Answer

Requests are handed to servers in turn, cycling through the list

Answer

Each request goes to the server with the fewest active connections

Answer

Requests are distributed by hashing the client IP address

Answer

All requests are sent to the fastest server

Question 13

What does a reverse proxy like Nginx typically do?

Accepted Answer

Sits in front of backend servers and forwards requests to them

Answer

Sends requests to the internet on behalf of clients to hide their IP

Answer

Optimizes and rewrites database queries

Answer

Converts backend code into static files

Question 14

What role does an API Gateway play in a microservices architecture?

Accepted Answer

A single entry point for clients handling routing, auth, and rate limiting

Answer

A shared database used by all services

Answer

A tool for running unit tests in CI

Answer

A queue that stores messages between services

Question 15

What is an endpoint like `/health` typically used for?

Accepted Answer

So a load balancer or orchestrator can tell if an instance is healthy

Answer

To return a list of all users in the system

Answer

To run database migrations at startup

Answer

To display Swagger documentation

Question 16

In Kubernetes, what is the difference between a liveness and a readiness probe?

Accepted Answer

Failed liveness restarts the container; failed readiness removes it from traffic

Answer

Failed liveness removes it from traffic; failed readiness restarts the container

Answer

Both do exactly the same thing with different names

Answer

Liveness runs only at build and readiness only at deploy

Question 17

According to the 12-Factor App, where should config like connection strings be stored?

Accepted Answer

In environment variables, separate from code

Answer

Hard-coded in the source code

Answer

In per-environment config files committed to the repository

Answer

Baked into the container image at build time

Question 18

What does the Processes factor of the 12-Factor App say?

Accepted Answer

Processes should be stateless and share-nothing

Answer

Each app must run as exactly one process

Answer

Processes should keep user sessions in memory

Answer

Processes should write persistent data to local disk

Question 19

What is the main difference between synchronous (REST/gRPC) and asynchronous (events) service communication?

Accepted Answer

In sync the caller waits for a reply; in async it publishes a message and moves on

Answer

Sync only uses UDP and async only uses TCP

Answer

Async communication is always faster and error-free

Answer

Sync needs no network because it's in one process

Question 20

What transport and serialization format does gRPC use by default?

Accepted Answer

HTTP/2 and Protocol Buffers

Answer

HTTP/1.1 and JSON

Answer

WebSocket and XML

Answer

UDP and MessagePack

Question 21

Why should every network call to another service have a timeout?

Accepted Answer

So a slow dependency can't tie up threads and resources indefinitely

Answer

So requests get encrypted

Answer

Because responses can't be cached without a timeout

Answer

So the target service processes faster

Question 22

Why should retries use exponential backoff?

Accepted Answer

To give a struggling service time to recover instead of piling on load

Answer

So requests are sent faster than before

Answer

Because retries are impossible without backoff

Answer

To eliminate the need for timeouts

Question 23

What is the core idea of the Circuit Breaker pattern?

Accepted Answer

After repeated failures, fail calls to the broken service fast for a while

Answer

Cutting power to a server when it overheats

Answer

Splitting traffic between two versions of a service

Answer

Limiting each user's requests per minute

Question 24

What is an idempotent operation?

Accepted Answer

One where executing it multiple times has the same effect as once

Answer

One that never returns an error

Answer

One that only reads data

Answer

One that runs in under one millisecond

Question 25

What is structured logging?

Accepted Answer

Writing logs in a parseable format like JSON with named fields

Answer

Writing logs in uppercase for readability

Answer

Storing logs in a separate table per day

Answer

Removing all logs in production

Question 26

What are the three pillars of observability?

Accepted Answer

Logs, metrics, and traces

Answer

CPU, RAM, and disk

Answer

Unit, integration, and e2e

Answer

Dev, staging, and production

Question 27

What is a correlation ID (or request ID) used for?

Accepted Answer

Linking all logs of a single request across different services

Answer

Uniquely identifying each user in the database

Answer

Encrypting the request body

Answer

Setting a request's priority in a queue

Question 28

What is the difference between SLI, SLO, and SLA?

Accepted Answer

SLI is the measurement, SLO the internal target, SLA the contractual promise to customers

Answer

SLA is the measurement, SLI the internal target, SLO the contract

Answer

All three are synonyms and interchangeable

Answer

SLI is for security, SLO for performance, SLA for cost

Question 29

According to the test pyramid, which type of test should be most numerous?

Accepted Answer

Unit tests

Answer

End-to-end tests

Answer

Manual tests

Answer

Load tests

Question 30

What is the difference between a unit test and an integration test?

Accepted Answer

Unit tests one piece in isolation; integration tests real components working together

Answer

Unit runs only in production and integration only in CI

Answer

Unit uses the UI and integration doesn't

Answer

No difference; only the test tool differs

Question 31

What is the difference between a mock and a stub in testing?

Accepted Answer

A stub returns canned answers; a mock verifies interactions (calls made)

Answer

A mock returns canned answers; a stub verifies calls

Answer

A stub is a full, lightweight working implementation

Answer

There is no difference

Question 32

What is a feature flag used for?

Accepted Answer

Turning a feature on/off at runtime without a new deploy

Answer

Marking bugs in the issue tracker

Answer

Choosing each service's programming language

Answer

Encrypting sensitive user attributes

Question 33

What happens in a blue-green deployment?

Accepted Answer

Two full environments exist and traffic switches from old to new at once

Answer

Instances are replaced one by one with the new version

Answer

The new version is given to only 1% of users

Answer

Code is edited directly on the production server

Question 34

What is a canary release?

Accepted Answer

Releasing to a small percentage of traffic and ramping up if healthy

Answer

Deploying the new version to all servers at once

Answer

Running tests only in staging

Answer

Keeping two full environments and switching at once

Question 35

What happens in a rolling deployment?

Accepted Answer

Instances are gradually replaced with the new version in batches

Answer

The whole system is shut down then started with the new version

Answer

Two complete parallel environments are maintained

Answer

The new version is enabled only for internal staff

Question 36

In DDD, what is the difference between an Entity and a Value Object?

Accepted Answer

An Entity has a stable identity; a Value Object is defined by its values and is immutable

Answer

An Entity is always immutable and a Value Object is mutable

Answer

A Value Object must have its own database table

Answer

Entities are only used in the controller layer

Question 37

What does Ubiquitous Language mean in DDD?

Accepted Answer

A shared vocabulary between developers and domain experts, also used in code

Answer

A single programming language for all services

Answer

Translating the UI into all languages

Answer

A standard JSON format for all APIs

Question 38

What is a Bounded Context in DDD?

Accepted Answer

A boundary within which a domain model and its terms have one consistent meaning

Answer

A container's memory limit

Answer

The maximum number of requests per user

Answer

The scope of variables in a function

Question 39

What is the main goal of Hexagonal Architecture (Ports & Adapters)?

Accepted Answer

Isolating domain logic from external details like databases and HTTP

Answer

Splitting the system into exactly six services

Answer

Putting all code in one file for simplicity

Answer

Replacing relational databases with NoSQL

Question 40

What problem does service discovery solve?

Accepted Answer

Finding the current addresses of service instances that keep changing

Answer

Finding bugs in service code

Answer

Discovering new services in a cloud marketplace

Answer

Automatically generating API documentation

Question 41

When is it better to hand off work like sending email to a background worker?

Accepted Answer

When the work is slow and the user doesn't need its result in the response

Answer

When the work must be atomic within the same database transaction

Answer

When the API response is meaningless without its result

Answer

Only when the server has no free CPU

Question 42

What does cold start mean in serverless?

Accepted Answer

Extra latency when a new execution environment is created and initialized

Answer

A function error when the data center is cold

Answer

The first deployment of a function in a new region

Answer

The function shutting down after every request

Question 43

Why is rate limiting often done at the edge (gateway or reverse proxy)?

Accepted Answer

So excess traffic is rejected before it reaches internal services

Answer

Because internal services can't return HTTP statuses

Answer

Because rate limiting is only possible with Nginx

Answer

So requests get encrypted

Question 44

What are sticky sessions on a load balancer?

Accepted Answer

A user's requests are always routed to the same instance

Answer

User sessions never expire

Answer

Sessions are stored in a shared database

Answer

Each request goes to a random instance

Question 45

What does graceful degradation mean in a backend system?

Accepted Answer

When one part fails, the system keeps working with reduced functionality

Answer

Gradually downgrading software to an older version

Answer

Shutting down entirely to prevent further errors

Answer

Lowering code quality to ship faster

Question 46

Which operation is typically the slowest?

Accepted Answer

A network round trip between continents

Answer

Reading from RAM

Answer

Accessing the CPU's L1 cache

Answer

A network round trip within the same data center

Question 47

What does a p99 latency of 800ms mean?

Accepted Answer

99% of requests completed in 800ms or less

Answer

The average response time is 800ms

Answer

99 requests are processed every 800ms

Answer

1% of requests completed in 800ms or less

Question 48

What is a modular monolith?

Accepted Answer

A single deployable with internal modules that have clear boundaries and interfaces

Answer

A set of microservices sharing one database

Answer

A monolith where each module deploys separately

Answer

A monolith that runs only on one physical server

Question 49

During a network partition, what choice does the CAP theorem force?

Accepted Answer

Between Consistency and Availability

Answer

Between Cost and Performance

Answer

Between Caching and Persistence

Answer

Between Concurrency and Atomicity

Question 50

What does eventual consistency mean?

Accepted Answer

If no new writes occur, all replicas eventually converge to the same value

Answer

Data is never synchronized between replicas

Answer

Every read always returns the latest write

Answer

Data is synchronized only once a day