System Design Explained: APIs, Databases, Caching, CDNs, Load Balancing & Production Infra

System Design · Days 31–40 · 2 min read

Video Reference: System Design Explained: APIs, Databases, Caching, CDNs, Load Balancing & Production Infra

1. Asli Problem Kya Hai? (The Core Why)

Socho ki Swiggy ka order system hai. Din mein 50 lakh orders aa sakte hain, har ek order ko 2‑3 second ke andar confirm karna hota hai. Agar APIs slow ho, database locks ho, ya CDN miss kare, toh customers ka order fail ho jayega – aur user trust kho jayega. Isi liye high‑traffic Indian products ko ek scalable, low‑latency architecture chahiye jo latency, throughput, aur availability ko balance kare.

2. Core Architecture & Mechanisms

2.1 APIs & Load Balancers

APIs ko front‑end se alag layer (API Gateway) mein rakhte hain. Load Balancer (LB) traffic ko multiple instances pe distribute karta hai, jisse throughput badhta hai aur single point of failure kam hota hai.

2.2 Databases – SQL vs NoSQL

• SQL: ACID compliance, strong consistency, ideal for transactional data (e.g., order details).
• NoSQL: eventual consistency, high write throughput, useful for session data, caching, analytics.

2.3 Caching

• In‑memory cache (Redis, Memcached) reduces DB hits, lowers latency.
• Read replicas: read traffic ko offload karte hain, read latency aur scalability improve hoti hai.

2.4 CDNs

Content Delivery Networks edge servers pe static assets (images, JS) store karte hain. Latency reduce hoti hai aur global reach achieve hota hai.

2.5 Production Infra

• Microservices: each service independent scaling.
• Service discovery & Gossip protocol: services automatically apne state share karte hain, fault tolerance badhata hai.
• Health checks & auto‑scaling: system ko dynamically adapt karne ke liye.

3. Architecture Flow

graph TD
    Client -->|HTTPS| API_Gateway
    API_Gateway --> LB
    LB --> Service_A
    LB --> Service_B
    Service_A -->|Read/Write| DB
    Service_B -->|Read/Write| DB
    DB -->|Replica| Read_Replica
    Service_A -->|Cache| Redis
    Service_B -->|Cache| Redis
    Service_A --> CDN
    CDN --> Client

4. Production Case Study

Zerodha:
Zerodha ne apni trading platform ke liye Redis as a cache layer use kiya, jisse order book updates milliseconds mein client ko mile. They also use Kafka for event streaming between services, ensuring low latency. For global clients, they deploy CDN (Akamai) for static assets. Load balancers (NGINX) distribute traffic across multiple geo‑regions, while read replicas handle heavy read traffic during market hours.

5. Trade-offs (Pros vs Cons)

6. System Design Interview Cheat Sheet

• Start with requirements: traffic, latency, consistency.
• Sketch high‑level flow: client → API → LB → services → DB/CDN.
• Choose storage: SQL for transactions, NoSQL for high‑write.
• Add caching: Redis/Memcached + read replicas.
• Plan scaling: autoscale, sharding, geo‑distribution.