本文轉載自 LinkedIn 作者 Arslan Ahmad 的 guiding principles to help you answer system design questions

有時間我也會再用中文補充的更詳細

Golden Rules in System Design

1. For a Read-Heavy System

• Consider using a Cache.

2. For a Write-Heavy System

• Use Message Queues for async processing

3. For a Low Latency Requirement

• Consider using a Cache and CDN.

4. Need 𝐀tomicity, 𝐂onsistency, 𝐈solation, 𝐃urability Compliant DB

• Go for RDBMS/SQL Database.

5. Have unstructured data

• Go for NoSQL Database

6. Have Complex Data (Videos, Images, Files)

• Go for Blob/Object storage.

7. Complex Pre-computation

• Use Message Queue & Cache.

8. High-Volume Data Search

• Consider search index, tries or search engine.

9. Scaling SQL Database

• Implement Database Sharding.

10. High Availability, Performance, & Throughput

• Use a Load Balancer.

11. Global Data Delivery

• Consider using a CDN.

12. Graph Data (data with nodes, edges, and relationships)

• Utilize Graph Database.

13. Scaling Various Components

• Implement Horizontal Scaling

14. High-Performing Database Queries

• Use Database Indexes

15. Bulk Job Processing

• Consider Batch Processing & Message Queues.

16. Server Load Management & Preventing DOS Attacks

• Use a Rate Limiter.

17. Microservices Architecture

• Use an API Gateway.

18. For Single Point of Failure

• Implement Redundancy.

19. For Fault-Tolerance and Durability

• Implement Data Replication.

20. For User-to-User fast communication

• Use Websockets.

21. Failure Detection in Distributed Systems

• Implement a Heartbeat.

22. Data Integrity

• Use Checksum Algorithm.

23. Efficient Server Scaling

• Implement Consistent Hashing.

24. Decentralized Data Transfer

• Consider Gossip Protocol.

25. Location-Based Functionality

• Use Quadtree, Geohash, etc.

26. Avoid Specific Technology Names

• Use generic terms.

27. High Availability and Consistency Trade-Off

• Eventual Consistency.

28. For IP resolution & Domain Name Query

• Mention DNS.

29. Handling Large Data in Network Requests

• Implement Pagination.

30. Cache Eviction Policy

• Preferred is LRU (Least Recently Used) Cache.

31. To handle traffic spikes

• Implement Autoscaling to manage resources dynamically

32. Need analytics and audit trails

• Consider using data lakes or append-only databases

33. Handling Large-Scale Simultaneous Connections

• Use Connection Pooling and consider using Protobuf to minimize data payload

34. To prevent overloading systems

• Use Token Bucked and Leaky Bucket algorithms

35. Minimize deployment downtime

• Implementing Blue-Green deployment to minimize downtime and reduce risks