System Design – Part 2: SNAKE

If we take Netflix as an example for SNAKE princile.

Scenario:

Step 1: Enumerate

• Register/Login
• Play movie
• Movie recommendation

Step 2: Prioritize

• Play movie:
  • Get channels
  • Get movies in channels
  • Play a movie in channel

Necessary:

Step 1: Ask

What’s the target DAU? For example: 5 millions

Step 2: Predict

User:

• Average concurrent users = daily active users / daily seconds * average oneline time
  • Average concurrent user = 5,000,000 / (24*60*60) * (30*60) = 104,167
• Peak users = average concurrent users * 6 = 625,000
• Max peak users in 3 months = peak users * 2 = 1,250,000

Traffic:

• Traffic per user = 3mpbs
• Max peak traffic 1,250,000 * 3mpbs = 3.75Tb/s

Memory:

• Memory per user = 10KB
• Max daily memory = 5,000,000 * 2 * 10 = 100GB

Storage:

• Total movie = 14,000
• Movie storage = total movie * average movie size = 14,000 * 50 GB = 700TB
  • Each movie may have different resolutions, so total can be counted around like 50GB

Application: serivce/algorithm

• Step 1: Replay the case, add a service for each request
• Step 2: Merge the services

Kilobit: data

• Step 1: Append dataset for each request below a service
• Step 2: Choose storage types

Evolve:

• Step 1: Analyze
  • with
    • Better: constrains – catch up new constrains
    • Broader: new cases – implement new use cases
    • Deeper: details – better performance and details
  • from the views of
    • Performance
    • Scalability
    • Robustness
• Step 2: Go back by evolving accordingly