Logo
Published on

CAP Theorem and PACELC - FAANG System Design Interview Guide

Table of Contents

🚀 CAP THEOREM (INTERVIEW CORE)

1. 🧠 What is CAP (REAL MEANING)

👉 You CANNOT guarantee all 3:

  • Consistency (C) → same data everywhere
  • Availability (A) → always respond
  • Partition Tolerance (P) → survive network failures

🧠 Analogy:

  • 2 WhatsApp servers (US + EU)
  • Network breaks ❌ Now you must choose:
  • Show latest data (C) ❌ OR
  • Always respond (A) ❌

👉 You can't do both.

🎯 FAANG Question

Q: Can a system guarantee C + A + P?

Answer: 👉 No 👉 In partition → must choose Consistency OR Availability

⚠️ Missing Insight (VERY IMPORTANT)

👉 P is NOT optional

  • In real world → network failures ALWAYS happen 👉 So real choice is:

C vs A (WHEN partition happens)

2. ⚖️ Consistency (C)

👉 All nodes return same latest data

🧠 Example:

  • Bank balance = always correct

🎯 FAANG Question

Q: What does strong consistency mean?

Answer: 👉 After write → all reads return updated value

⚠️ Missing Insight

  • Includes:
    • linearizability
    • read-after-write consistency

3. ⚡ Availability (A)

👉 System always responds

🧠 Example:

  • Instagram loads even if slightly outdated

🎯 FAANG Question

Q: Is stale data acceptable?

Answer: 👉 Depends:

  • Social media → YES
  • Banking → NO

⚠️ Missing Insight

  • Availability ≠ correctness
  • It only guarantees response

4. 🌐 Partition Tolerance (P)

👉 System continues despite network failure

🧠 Example:

  • Data centers can't talk but system still runs

🎯 FAANG Question

Q: Can we ignore partition tolerance?

Answer: 👉 No. 👉 Distributed systems MUST handle partitions.

🚀 CAP COMBINATIONS (IMPORTANT)

5. 🔵 CP Systems (Consistency + Partition)

👉 Sacrifice availability

🧠 Example:

  • Bank system refuses request if unsure

🎯 FAANG Question

Q: When choose CP?

Answer: 👉 When correctness > uptime 👉 e.g., payments, inventory

6. 🟢 AP Systems (Availability + Partition)

👉 Sacrifice consistency

🧠 Example:

  • Social media feed

🎯 FAANG Question

Q: When choose AP?

Answer: 👉 When uptime > correctness 👉 e.g., likes, comments

7. ❌ CA (Myth)

👉 Not practical

🎯 FAANG Question

Q: Why CA doesn't exist?

Answer: 👉 Because partition will happen. 👉 System forced into C or A.

🚀 REAL DATABASE MAPPING

SQL → CP

  • Strong consistency
  • May reject requests

NoSQL → AP

  • High availability
  • Eventual consistency

🎯 FAANG Question

Q: Why NoSQL is AP?

Answer: 👉 Prioritizes uptime + scaling. 👉 Accepts temporary inconsistency.

🚀 PACELC THEOREM (ADVANCED 🔥)

8. 🧠 Why CAP is incomplete?

👉 CAP only talks about failure case

👉 But:

What happens when system is NORMAL?

9. ⚡ PACELC Definition

👉 Two decisions:

When Partition (P):

  • Choose A or C

Else (E) (No partition):

  • Choose Latency (L) or Consistency (C)

🧠 Analogy: Even when network is fine:

  • Wait for all replicas (slow but consistent)
  • OR respond fast (may be slightly stale)

🎯 FAANG Question

Q: What does PACELC add over CAP?

Answer: 👉 Adds latency vs consistency tradeoff in normal state.

10. ⚖️ Latency vs Consistency

Low Latency

  • Fast response
  • May read stale data

High Consistency

  • Slower (sync replication)

🎯 FAANG Question

Q: Why consistency increases latency?

Answer: 👉 Need coordination across nodes. 👉 More network round trips.

🚀 REAL SYSTEM CLASSIFICATION

11. 🟢 PA/EL Systems

👉 Choose:

  • Partition → Availability
  • Normal → Low latency

Examples:

  • Apache Cassandra
  • Amazon DynamoDB

🎯 FAANG Question

Q: Why Cassandra is PA/EL?

Answer: 👉 Always available. 👉 Fast reads/writes. 👉 Accepts eventual consistency.

12. 🔵 PC/EC Systems

👉 Choose:

  • Partition → Consistency
  • Normal → Consistency over latency

Examples:

  • HBase
  • Bigtable

🎯 FAANG Question

Q: Why choose PC/EC?

Answer: 👉 Critical correctness systems. 👉 Accept slower responses.

13. 🟡 Hybrid (Configurable)

Example:

  • MongoDB

🎯 FAANG Question

Q: How MongoDB behaves?

Answer:

  • Default → PA/EC
  • With majority writes → PC/EC

⚠️ Missing Insight

👉 Many modern DBs are tunable

  • You can adjust consistency level

🚀 MOST IMPORTANT INTERVIEW INSIGHTS

14. 🔥 CAP IS ABOUT TRADE-OFFS

👉 Not about what system "is" 👉 About what system chooses

🎯 FAANG Question

Q: Is a system always CP or AP?

Answer: 👉 No. 👉 Behavior depends on configuration & situation.

15. 🧠 Eventual Consistency Reality

👉 System becomes consistent over time

🎯 FAANG Question

Q: Is eventual consistency safe?

Answer: 👉 Yes for: Social apps. 👉 No for: Financial systems.

16. ⚡ Hidden Concept: Read/Write Quorum

👉 Helps balance C vs A

🎯 FAANG Question

Q: How to improve consistency in AP system?

Answer: 👉 Use quorum:

  • Read from majority
  • Write to majority

17. 🔁 Replication Impact

👉 More replicas = more availability 👉 But: 👉 More coordination = more latency

🎯 FAANG Question

Q: Why replication increases latency?

Answer: 👉 Writes must sync across nodes

18. 🚀 Real-World Design Thinking

🎯 FAANG Question

Q: Design WhatsApp messages consistency?

Answer:

  • Sent → must be durable (C)
  • Seen status → can be eventual (A)

👉 Mix of CAP choices

19. 🧩 Common Mistake

👉 "NoSQL = always eventual consistency" ❌

🎯 FAANG Question

Q: Can NoSQL be strongly consistent?

Answer: 👉 Yes (configurable)

20. 🔥 Ultimate Interview Answer

🎯 FAANG Question

Q: Explain CAP in one line

Answer: 👉 "In a distributed system, during network partition, you must choose between consistency and availability."

🚀 FINAL MEMORY HOOK

👉 CAP = Failure tradeoff (C vs A) 👉 PACELC = Always tradeoff (Latency vs Consistency)

🔥 ULTRA-SHORT SUMMARY

  • CAP → happens during failure
  • PACELC → happens ALWAYS
  • Real systems → configurable + hybrid