Question 1

What sits at the heart of every system, and what is system design trying to do with it?

Accepted Answer

Data. WhatsApp, YouTube, LinkedIn, Instagram all present data to users as text, images, audio or video. System design is about finding efficient ways to store, process and deliver that data in the form users expect.

Question 2

List the core components of a system and give each one a one-line responsibility.

Accepted Answer

- Database: Persistent storage; store and retrieve data reliably. SQL (MySQL, PostgreSQL) or NoSQL (MongoDB, Cassandra). Drawn as a cylinder.
- Application code (backend): Business logic; exposes APIs; performs insert, update, retrieve, delete against the database.
- Client application: User-facing interface (mobile, web, desktop); takes input, calls APIs, renders JSON as UI.
- Cache: Fast-access layer for frequently requested data; cuts database load and latency.
- Load balancer: Spreads traffic across servers; checks which servers are healthy.
- Message queue: Broker between services for asynchronous work; retries, buffering.
- Monitoring and logs: Health of every component; error detection; stack traces for debugging.

Question 3

What does the backend (application code) actually do, and how does it talk to the database?

Accepted Answer

It is the middle layer that hides tables, documents and queries from the user. It exposes endpoints such as /getUser, /uploadVideo, /createOrder, /deletePost, handles insert, update, retrieve and delete, and communicates with the database on the user's behalf. Connection to the database is by IP address over protocols like TCP or HTTP. Users never write queries; they hit an API and the application does the rest.

Question 4

Why is JSON the default data format for APIs?

Accepted Answer

JSON (JavaScript Object Notation) is a standardized, lightweight, human-readable key-value structure that every language can parse. The API sends and receives JSON; the client turns it into a visual, user-friendly format.

Question 5

What is the client application responsible for, and how does it reach the backend?

Accepted Answer

Display data, take user input, send requests to backend APIs, show the responses. Examples: mobile apps, browsers, desktop apps. It communicates with the backend over HTTP, HTTPS or TCP/IP.

Question 6

What is a cache and why put one in front of the database?

Accepted Answer

A cache stores frequently accessed data in a fast-access layer so the system does not query the database again and again. It reduces database load, speeds up responses and holds temporary, hot data. Example: thousands of users request the same homepage feed; the cache serves it directly instead of hitting the database each time.

Question 7

What are the two responsibilities of a load balancer?

Accepted Answer

1. Health: check whether each server is active and responsive.
2. Distribution: divide traffic across the available servers so no single one is overwhelmed.

Question 8

Synchronous vs asynchronous communication, using the order-service example.

Accepted Answer

- Synchronous: Definition - Caller waits for the response before proceeding. Example - Payment confirmation before the order completes.
- Asynchronous: Definition - Caller sends the request and moves on. Example - SMS notification after the order is placed.

When an order is placed, inventory must update, delivery must be notified, and email/SMS must go out. Doing all of that in sequence makes the response slow, so the non-critical work is pushed to a message queue.

Question 9

How does a message queue take work off the order service? Give the steps.

Accepted Answer

1. The order service pushes requests into the queue.
2. The queue stores and manages them independently.
3. Connected services (SMS, email) pick requests up and process them one by one.
4. The queue retries a request if it fails.
5. It may send a batch summary (fulfilled / unfulfilled) back to the order service at the end of the day.

The order service is no longer responsible for tracking individual notifications; the queue owns that.

Question 10

Why are monitoring and logs part of system design rather than an afterthought?

Accepted Answer

Monitoring confirms every component is working; logs give the trail of events to debug. Concretely you need them to verify the load balancer is distributing traffic, detect errors in application code, read the full stack trace, reproduce the issue locally, and confirm a new deployment did not break existing features. Key activities: health checks, error detection and alerting, performance tracking, impact analysis of new features.

Question 11

What is latency?

Accepted Answer

The time a system takes to respond to a request: the response delay, waiting time or processing delay. When a user taps "Open Instagram feed" or "Play YouTube video", the time until the result appears is latency.

Question 12

What is a data-intensive application? Give examples.

Accepted Answer

One where the main challenge is handling large amounts of data efficiently: storing, retrieving, transferring, updating, synchronizing and serving it quickly.

- Instagram feed: Millions of posts served to millions of users at once
- WhatsApp messages: Constant reads and writes across devices
- Banking transactions: High volume of transactional reads and writes
- Analytics dashboard: Aggregating and displaying large datasets in real time
- Log processing: Ingesting, storing and querying massive log files

Question 13

What performance issues and key concerns show up in data-intensive systems?

Accepted Answer

Issues: slow database responses, bulky network calls that move too much data, under-sized server configuration.

Concerns to design for: how fast can we read? how safely do we store? how many users can access simultaneously? what if a server dies? what if a network call breaks?

Question 14

Which techniques address data-intensive problems?

Accepted Answer

- Caching: Keep hot data close to the user for fast reads
- Sharding: Split data across multiple databases to spread load
- Replication: Keep copies on several servers for fault tolerance
- Consistency models: Keep data accurate across a distributed system
- Database optimization: Indexing, query tuning, choosing the right database type

Question 15

What is a compute-intensive application, and what are its key concerns?

Accepted Answer

The bottleneck is processing power (CPU/GPU): the data fetched may be small, but transforming it is heavy. Examples: image processing (pixel-level filters, resizing), video rendering (frame-by-frame encode/decode), ML model training, physics or financial simulations, cryptography.

Concerns: how fast can we compute? can we parallelize across cores or machines? can we reduce algorithmic complexity? should we use GPU instead of CPU?

Question 16

Compare data-intensive and compute-intensive applications.

Accepted Answer

- Bottleneck: Data-intensive - Data movement and storage. Compute-intensive - Processing power.
- Focus: Data-intensive - Fast retrieval and delivery. Compute-intensive - Fast computation.
- Hardware concern: Data-intensive - Disk I/O, network, memory. Compute-intensive - CPU / GPU.
- Data volume: Data-intensive - Very high. Compute-intensive - Relatively low.
- Processing complexity: Data-intensive - Low to moderate. Compute-intensive - Very high.
- Solutions: Data-intensive - Caching, sharding, replication. Compute-intensive - Parallelism, GPU acceleration, algorithm optimization.

Question 17

Give the rule of thumb, and show how YouTube or Instagram is both at once.

Accepted Answer

If time is lost in data movement, it is data-intensive; if time is lost in computation, it is compute-intensive. Most real systems are both, feature by feature:

- YouTube: storing, streaming and delivering video is data-intensive; the recommendation engine (AI models, analytics, personalization) is compute-intensive.
- Instagram: feed loading, messages and media delivery are data-intensive; image filters, face detection and AI recommendations are compute-intensive.

Identifying which bottleneck you have decides the architecture, the scaling strategy and the optimization techniques.

Question 18

Define functional and non-functional requirements with an e-commerce example.

Accepted Answer

Functional = what the system does (features): register, log in, browse, search, filter, add to cart, apply coupon, place order, pay, track order.

Non-functional = how well it does it (quality attributes): handle 1 million concurrent users, response time under 200 ms, 99.9% availability, data encrypted, scalable to 10x traffic, fault-tolerant.

Both matter equally: a system with every feature that crashes under load is as unusable as a fast system missing core features.

Question 19

Why do functional requirements matter specifically to the engineer?

Accepted Answer

They tell you what to build: which APIs are needed, which database operations each feature requires, and how users will interact with the system. The endpoint list and the schema fall out of the functional requirements.

Question 20

List the seven key non-functional attributes.

Accepted Answer

1. Scalability: handle increasing load without degrading (vertical or horizontal).
2. Availability: percentage of time the system is up and reachable.
3. Reliability: correct results and no data loss, even during failures.
4. Performance (latency): how fast it responds, measured in percentiles.
5. Security: authentication, authorization, encryption.
6. Maintainability: easy to modify and extend; modular code.
7. Observability: understand internal state from outputs; monitoring plus logging.

Fundamentals

Fundamentals