Question 1

What is idempotency and why does it matter?

Accepted Answer

**Target Audience:** Beginners who have built or called a web API and want to understand why retries are dangerous and how real systems make them safe.
**Difficulty:** Medium
**Verified with:** Node 22.22 and Python 3.12 (both print identical output). HTTP method semantics are cited from RFC 9110; Stripe behaviour from the Stripe API reference ("Idempotent requests"); the Idempotency-Key header from the IETF httpapi working-group Internet-Draft.

> **How to read this doc:** Every idea is explain

Question 2

How do you handle asynchronism, message queues and backpressure at scale?

Accepted Answer

## Step 1: Use Cases, Constraints, Assumptions, Back-of-Envelope

### Use cases
We'll scope to handle only the core flows that expose the bottlenecks; everything else is deferred.

- **User** triggers async work — POST returns 202, job enqueued (email, thumbnail, analytics), polled or webhook on done.
- **Service** consumes — workers pull with prefetch, ack on success, DLQ on poison, retry with backoff.
- **Service** handles backpressure — bounded queues, shed/slow producers, scale consumers on

Question 3

How would you design OpenTelemetry tracing for 200 microservices?

Accepted Answer

**Target Audience:** Engineers who know what a trace is (see the pillars doc) and want to design tracing for a real, polyglot microservice fleet, and candidates preparing for an observability design round.
**Difficulty:** Medium
**Verified with:** Node 22.22 and Python 3.12 (both print identical output). Header format from the W3C Trace Context Recommendation; Collector component behaviour from the loadbalancing exporter README on pkg.go.dev and Grafana Alloy documentation of the tail sampling p

Question 4

Kafka vs SQS — how do you choose for an event-driven system in 2025?

Accepted Answer

**Target Audience:** Beginners to messaging — you know what a server and a database are; queues, logs, offsets and consumer groups are defined as we go.
**Difficulty:** Medium
**Verified with:** Node 22.22 and Python 3.12 (both print identical output). Product limits are from the AWS SQS Developer Guide and the Apache Kafka release announcements, checked in October 2026.

> **How to read this doc:** Every idea is explained in plain language first, then tagged with `📌 Interview term:` so you als

Question 5

What are blue-green and canary deployments?

Accepted Answer

**Target Audience:** Beginners to release engineering — you know what a server, a load balancer and a deploy are; every term is defined as we go.
**Difficulty:** Medium
**Verified with:** Node 22.22 and Python 3.12 (both print identical output). Kubernetes defaults are from the Kubernetes Deployment API reference; the canary definition is from the Google SRE Workbook, chapter 16, checked in October 2026.

> **How to read this doc:** Every idea is explained in plain language first, then tagged wi

Question 6

How would you design a distributed job scheduler?

Accepted Answer

**Target Audience:** Engineers preparing for a system design round. You know what a database, a queue and a server are; leases, fencing and misfire policies are explained from zero.
**Difficulty:** Hard
**Verified with:** Node 22 and Python 3.12 (identical output). Kubernetes CronJob and PostgreSQL facts are cited inline, checked in October 2026.

> **How to read this doc:** Plain language first, then `📌 Interview term:` callouts with the vocabulary an interviewer expects. Every number in the d

Question 7

How would you design a notification system?

Accepted Answer

**Target Audience:** Engineers preparing for a system design round; queues, retries and push notifications are explained from the start.
**Difficulty:** Hard
**Verified with:** Node 22.22 and Python 3.12 (both print identical output). Provider facts (FCM HTTP v1 migration, FCM payload size, SMS segment lengths, Gmail and Yahoo bulk sender rules) checked in October 2026 and cited inline.

> **How to read this doc:** Every idea is explained in plain language first, then tagged with `📌 Interview t

Question 8

How do you scale WebSockets to 1M connections with Redis Pub/Sub?

Accepted Answer

**Target Audience:** Engineers who have run a single WebSocket server and now need many. Read [When would you use WebSockets vs polling vs SSE?](/interview-question/when-would-you-use-websockets-vs-polling-vs-sse) and [WebSocket presence, sticky sessions, and resume — do you need sticky LB?](/interview-question/websocket-presence-sticky-sessions-and-resume-do-you-need-sticky-lb) first if those are new.
**Difficulty:** Hard
**Verified with:** Node 22.22, Python 3.12, Go 1.24 and Java 21 (all four

Question 9

How would you design AI inference infra with GPU queue and KV-cache?

Accepted Answer

**Target Audience:** Engineers preparing for a system design round who may never have served a model. GPUs, tokens, prefill, decode and the KV-cache are explained from zero.
**Difficulty:** Hard
**Verified with:** Node 22 and Python 3.12 (identical output). Paper and product facts (Orca, vLLM, model shapes, GPU memory) are cited inline, checked in October 2026; throughput per node is an **assumption**, stated as such.

> **How to read this doc:** Plain language first, then `📌 Interview term:` c

Question 10

Exactly-once processing — can Kafka or SQS guarantee it?

Accepted Answer

**Target Audience:** Engineers who have used a queue or Kafka and heard "exactly-once" promised; everything from offsets to transactions is defined as we go.
**Difficulty:** Hard
**Verified with:** Node 22.22 and Python 3.12 (both print identical output). Kafka and SQS behaviour is from the Apache Kafka KIPs and documentation and the AWS SQS Developer Guide, checked in October 2026.

> **How to read this doc:** Every idea is explained in plain language first, then tagged with `📌 Interview term:

SYSTEM DESIGNDECODED.

Medium