System Design Fundamentals

# Why System Design?

As a Senior Engineer, you're expected to design systems, not just write code. System design is about making trade-offs — there's no perfect solution, only the best fit for your constraints.

# Load Balancing

A load balancer sits between clients and servers, distributing traffic to prevent any single server from becoming a bottleneck.

Client Requests
       │
       ▼
┌─────────────┐
│ Load Balancer│
└──────┬──────┘
    ┌──┴──┬──────┐
    ▼     ▼      ▼
 Server  Server  Server
   A       B       C

Key Strategies

• Round Robin — rotate through servers sequentially
• Least Connections — route to the server with fewest active requests
• IP Hash — same client IP always hits the same server (sticky sessions)
• Weighted — more powerful servers get more traffic

# Caching

Caching stores frequently accessed data in a faster layer (memory) to avoid expensive recomputation or database queries.

Cache Layers

Browser Cache → CDN Edge → App Cache (Redis) → Database
  (~0ms)       (~5ms)       (~1ms)              (~10-100ms)

Cache Patterns

• Cache-Aside — App checks cache first, fills on miss (most common)
• Write-Through — Writes go to cache AND database together
• Write-Behind — Writes go to cache only, async flush to DB (risky)
• Read-Through — Cache itself fetches from DB on miss

# CDN (Content Delivery Network)

CDNs distribute static content to edge servers worldwide. A user in Tokyo gets served from an Asian edge node instead of a US-based origin server — cutting latency from ~150ms to ~5ms.

Push vs Pull CDN

• Pull — CDN fetches from origin on first request, then caches. Simple, but first request is slow.
• Push — You upload content to CDN proactively. Better for content you know will be popular.

# Horizontal vs Vertical Scaling

• Vertical (Scale Up) — Add more CPU/RAM to one machine. Simple but has a ceiling and a single point of failure.
• Horizontal (Scale Out) — Add more machines. No ceiling, better fault tolerance, but adds complexity (load balancing, data consistency).

Most real-world systems use both: vertically scale each machine to a reasonable size, then horizontally scale the fleet.