Template — The 13-Step HLD Answer
In one line
The HLD interview skeleton: 13 steps that work for Twitter, Uber, payment systems, or a CI/CD platform. Run them in order, out loud.
Key ideas
- 01
In plain words: think of this like a recipe card. Whatever system you're asked to design — Twitter, Uber, a payment system — you follow the same 13 numbered steps, in the same order, every single time. You never have to wonder 'what do I say next?' again.
- 02
1 · Requirements — functional + non-functional, written down; ask the 8 questions (Phase 0).
- 03
2 · Scale estimation — QPS, peak, storage, bandwidth, memory (Phase 9.9 math).
- 04
3 · APIs — the surface: verbs, paths, request/response shapes, status codes.
- 05
4 · Data model — core tables/objects + access patterns + consistency budget.
- 06
5 · High-level architecture — the boxes: clients, LB, services, data stores, queues, caches, CDN.
- 07
6 · Core request flow — trace ONE request write path + ONE read path through the boxes.
- 08
7 · Bottlenecks — name the top 2 constraints (write hot spot, DB, bandwidth, and so on).
- 09
8 · Scaling solution — how each bottleneck gets relieved: reads (cache/replicas/CDN), writes (partition/queue), and so on.
- 10
9 · Failure handling — node death, dependency timeout, queue overflow, region failure (Phase 14.1–14.2).
- 11
10 · Consistency — per-path choice: strong/eventual, read-your-writes, outbox/saga if needed.
- 12
11 · Security — auth flow, mTLS, PII/tenancy, rate limits (Phase 14.4–14.6).
- 13
12 · Monitoring — the golden signals + trace/log correlation (Phase 14.3).
- 14
13 · Trade-offs — top 3 decisions and the explicit cost of each; what you'd reconsider.
Java / Spring map
- →
Take the template home: for any two real systems, run the whole 13-step script into a Notion page.
Code & diagrams
Print this. Tape it to a wall. It is the whole HLD game.
1. Requirements (functional + NFRs)
2. Scale estimation (QPS, peak, storage, bandwidth)
3. APIs (verbs, paths, status codes)
4. Data model (tables + access patterns + consistency)
5. High-level arch (draw the boxes)
6. Core request flow (one write path, one read path)
7. Bottlenecks (top 2, with numbers)
8. Scaling (read: cache/replica/CDN · write: partition/queue)
9. Failure handling (node, dependency, queue, region)
10. Consistency (per path: strong vs eventual + mechanisms)
11. Security (auth + mTLS + PII + rate limits)
12. Monitoring (golden signals + traces)
13. Trade-offs (top 3 decisions and their cost)Explain without notes
Without notes, run the 13 steps on 'design a notification platform' in 20 minutes.
Practice
Do two full practice rounds with a timer: one on Uber, one on CI/CD platform.
Trade-offs
- ↔
Skipping to architecture without scale numbers is the #1 deduction in HLD rounds — the template forces it.
Run it in production
You've designed it. Now build, operate, and break the same idea hands-on in the DevOps courses:
Completion checklist
I can deliver the 13-step skeleton from memory on any prompt.