Top 30 Scalability Interview Questions & Answers 2025

Vertical (scale up): add more power to existing machine (CPU, RAM, storage). Horizontal (scale out): add more machines. Vertical: simpler, limits exist, single point of failure. Horizontal: more complex, near-infinite scale, fault tolerant. Modern preference: horizontal for web services. Databases often start vertical, then shard for horizontal.

Methods: (1) Load testing (JMeter, k6, Locust), (2) Profiling (APM tools), (3) Metrics monitoring (CPU, memory, I/O, network), (4) Database query analysis, (5) Distributed tracing. Common bottlenecks: database (most common), CPU-bound operations, I/O (disk, network), external services. Address biggest bottleneck first (Amdahl's Law).

Strategies: (1) Caching (Redis, CDN), (2) Read replicas (distribute read load), (3) Database indexing (faster queries), (4) Denormalization (fewer joins), (5) Query optimization, (6) Connection pooling, (7) Content delivery networks, (8) Materialized views. Often combine strategies. Measure impact of each change.

Strategies: (1) Database sharding (distribute data), (2) Write-behind caching (async writes), (3) Message queues (buffer writes), (4) Event sourcing (append-only), (5) CQRS (separate read/write paths), (6) Batch writes, (7) Optimistic locking (reduce contention). Write scaling harder than read. Design for write patterns early.

Architecture: (1) Stateless services (easy horizontal scaling), (2) Load balancers (distribute traffic), (3) CDN for static content, (4) Caching layers (reduce DB load), (5) Database sharding, (6) Async processing (queues), (7) Microservices (scale independently), (8) Auto-scaling (handle load spikes). Test with realistic load. Plan for 10x growth.