Skip to main content

How to approach a System Design Problem

· 2 min read

Why​

  • Your question begins with “Why”, Problem description

Requirements​

  • functional requirements - hard requirements:
    • must achieve
  • non-functional requirements:
    • soft, but user experience oriented like latency, search speed, autocomplete…

Scale Estimation​

  • QPS (Queries Per Second)
  • RAM size
  • Database size
  • Read/Write loads

Database Schema​

  • What fields do we need
  • How many bytes per field
  • Total size per entry

System APIs​

  • Define Key APIs
  • What are the Parameters for each of the APIs
    • name
    • type
    • is it optional
    • What are the entries in the returned JSON

Algorithm And Data Structure Design (30% of the interview)

Talk about what algorithms are associated with this problem​

  • Evaluate the run time and memory usage
  • Compare different options and trade-offs (Bonus: a lot of interviewers love this)
  • Examples:
    • Hashing, MD5, SHA256, Base64 for URL encoding

Talk about what data structures to use for this problem​

  • Best data structure for different data type:
    • Segment Tree, Priority Queue for sliding window or data stream
    • Quad Tree for Spatial or Geometrical data points
    • Trie/Prefix tree for Autocomplete

Dev Operation/Site Reliability/Maintenance (30%)

Data Partitioning​

  • Sharding
  • Consistent Hashing

Replication and Fault Tolerance​

  • CAP Theorem
  • Dockers/Kubernetes

General Improvements/Further Information (20%)

Cache​

  • Redis, MemCache
  • Frontend Cache:
    • Talk about page stores like Mobx and Redux

Load Balancing / middleware to improve performance​

  • Message Queue/Bus, Apache Kafka
  • NgNix For routing

Business Model (How to generate money with this app?)​

  • VIP access
  • Ranking, Sorting Priority

Testing Strategy​

  • How would you test each component

CI/CD​

  • What’s the release cycle like