System Design Concepts for Beginners: Capacity Estimations

Introduction

• Capacity estimations in system design interviews are crucial, especially in big tech companies.
• Typically follows after clarifying system requirements.
• Often, estimations don’t directly affect design in interviews.
• Estimations demonstrate problem-solving approach.

Importance in Real Life

• To determine resources needed for high-level system functioning:
  • Number of servers
  • Data partitions
  • Memory for caching

Key Points of Estimation

• Demonstrates organized thinking and problem-solving skills.
• Important numbers to estimate:
  • Traffic
  • Storage
  • Bandwidth
  • Memory for caching
• Focus on traffic and storage in interviews; optional to include bandwidth and caching.

Traffic Estimation

• Input: Number of daily active users (DAUs).
  • Example: 10 million DAUs.
• Convert users to requests:
  • Consider user interactions (read and write requests).
• **Typical usage consideration: **
  • Example scenario: Paste bin service (upload & view text).
  • Assume 10% users upload once per day; therefore, 1 million write requests per day.
  • Determine read to write ratio; assume read-heavy (example: 50 to 1 ratio).
  • Calculate read requests per day (example: 50 million reads per day).
• Convert to requests per second:
  • Divide daily requests by seconds in a day (~100K seconds).
  • Example results: 10 write requests/second and 500 read requests/second.

Storage Estimation

• Identify data artifacts: Focus on the largest storage requirement.
  • Example: Pastes in a paste bin service.
• **Estimate storage based on data size: **
  • Example: Average paste size ~10KB.
  • Calculate daily storage need: 10KB * 1 million write requests = 10GB/day.
• **Consider data retention period: **
  • Example: Data expires in 5 years => Storage for 5 years.
  • Calculate total storage: 10GB/day * 2000 days = 20TB.
  • Consider data replication (example: 3 times) => 60TB total.

Bandwidth Estimation

• **Incoming data per second: **
  • Write requests: 10 requests/second * 10KB = 100KB/second.
• **Outgoing data per second: **
  • Read requests: 500 requests/second * 10KB = 5MB/second.

Cache Memory Estimation

• 80-20 Rule for Caching: 20% of data generating 80% of traffic.
  • Example: Cache 20% of read requests.
  • Calculate memory for cache: 50 million daily reads * 2KB = 100GB.
  • Consider duplicates; actual usage will be less.*

Application Servers Estimation

• Formula: Number of requests per second / Requests a server can handle.
  • Consider server specifications and if CPU-bound.
  • Example: 500 requests/second, 16 requests/server => 30-50 servers needed.

Common Sizes and Estimations

• **Text Estimates: **
  • English language ~500K words.
  • One line text ~10 words, one word ~5 characters or 5 bytes.
• **Media Estimates: **
  • HD image ~3MB (1280x720 pixels, 24-bit depth).
  • Profile image ~300KB (300x300 pixels).
  • 1 minute HD video ~50MB (compression included).
  • Consider storage for multiple resolutions (double size for HD).

Conclusion

• Estimations are approximate; aggressive approximation is key.
• Understand common sizes for text and media.
• Demonstrates organized and methodical problem-solving skills in design interviews.