System Design: NFRs, Critical Path, Architecture Patterns, LLD Framework

// System Design Concepts — NFRs, Scalability, Architecture Patterns

// NFR Examples for C++ / HFT systems:
struct NFR {
    // Latency:     < 1ms P99 (market data handler)
    // Throughput:  > 100K events/sec
    // Availability: 99.99% (52 min downtime/year)
    // Reliability:  zero message loss
};

// Critical Path (Trading):
// Market Data → Feed Handler → Strategy → Risk Check → Order Router → Exchange
// Every step on critical path must be < 1ms; zero blocking calls

// Latency vs Throughput tradeoff:
// Batching: group N requests → send together → HIGHER throughput, HIGHER latency
// Streaming: send each immediately → LOWER latency, LOWER throughput

// Architecture selection matrix:
// Low latency, tight control     → Monolith / in-process pipeline
// Audit trail, replay            → Event-driven / event sourcing
// Complex domain rules           → Clean / Hexagonal architecture
// Separate read/write perf       → CQRS
// Runtime swappable algorithms   → Strategy pattern
// Independent deployment scaling → Microservices (only if team is ready!)

// Pipeline pattern (C++ market data):
// Stage 1: Decode raw bytes → Stage 2: Normalize ticks →
// Stage 3: Strategy → Stage 4: Risk → Stage 5: Order
// Each stage: dedicated thread + SPSC lock-free queue between stages

// CQRS — separate read/write models:
// Command side: validates + writes (source of truth)
// Query side:   denormalized read model (fast reads, eventual consistency)
// Event bus synchronizes write→read side asynchronously

// Scalability:
// Vertical (scale-up): bigger machine; has ceiling; needs restart
// Horizontal (scale-out): more instances behind LB; stateless required

// LLD approach (5 steps):
// 1. Clarify requirements (functional + NFR)
// 2. Identify core classes / entities
// 3. Define interfaces (abstract, not concrete)
// 4. Apply design patterns where they fit naturally
// 5. Show extensibility: how to add new feature without breaking existing