TechPartner

← Back to Blog
System Design course for beginners — scale systems from 1 to 1M users · Part 1 / 10

System design for beginners (1/10): Metrics, QPS, and CAP theorem

What is system design? Free 10-part course: requirements, QPS, p99, availability, CAP—ShortLink traffic estimates from 1 to 1M users. Roadmap: learn-system-design-basics-from-scratch

· 1 min read

System Design course for beginners — scale systems from 1 to 1M users · Part 1/10
  1. 1Metrics & CAP foundations
  2. 2Single-server monolith
  3. 3DB split & ops
  4. 4Redis + CDN
  5. 5Read replicas
  6. 6Load balancing
  7. 7Message queues
  8. 8Sharding & regions
  9. 9Interview cheat sheet
  10. 10ShortLink capstone
Table of contents
  1. What a design session answers
  2. Metrics
  3. Back-of-envelope (1M MAU sketch)
  4. CAP (practical)
  5. Course roadmap
Lộ trình scale: 1 → 1.000.000 người dùng11001K10K100K500K1MMonolithManaged DBRedis + CDNReplica + LBQueueMỗi bài = thêm component khi metric vượt ngưỡng — không scale sớm
Evolution overview: each scale tier typically adds one layer (managed DB, cache, replica, LB, queue, shard).

System Design basics — scale from 1 → 1,000,000 users — Part 1/10
Running case study: ShortLink (URL shortener, redirect, click counts).

You need basic HTTP, APIs, and relational CRUD — not prior distributed-systems experience.


What a design session answers

  1. Clarify scope and v1 features
  2. Estimate traffic and storage
  3. Define APIs and core schema
  4. Draw a high-level diagram
  5. Find the first bottleneck
  6. Deep-dive only where it matters

Metrics

Design for peak QPS, track p99 latency, and set availability targets (e.g. 99.9%).

ShortLink is read-heavy: redirects dominate writes.


Back-of-envelope (1M MAU sketch)

~300k DAU × 5 redirects/day ≈ 1.5M redirects/day → ~17 RPS average; multiply by 10 for a conservative peak planning number.


CAP (practical)

Strong consistency for slug creation; eventual consistency is fine for aggregated click counts.


Course roadmap

Monolith → managed DB → cache/CDN → replicas → load balancing → queues → sharding/multi-region.

Next: Part 2 — single-user monolith


Series navigation: Part 2 →

Frequently asked questions

How is system design different from algorithm interviews?

Algorithms focus on single-machine complexity. System design focuses on multi-machine trade-offs: network, storage, latency, cost, and reliability.

Do I need Kubernetes first?

No. This course explains components conceptually; orchestration comes after you know why horizontal scaling is needed.

Does your startup need a tech partner?

Free chat about MVP, timeline, and the right profit-sharing approach.