Load Balancing

Strategies

• Round Robin simple, no intelligence
• Least Connections routes to server with fewest active connections
• Weighted factors in server capacity
• IP Hash consistent routing for session affinity

Layers

• L4 (Transport) TCP/UDP level, faster, less intelligent
• L7 (Application) HTTP level, can route based on path/headers

Caching

Cache Strategies

• Cache-Aside app checks cache first, falls back to DB
• Write-Through writes go to cache and DB simultaneously
• Write-Behind writes go to cache, async flush to DB
• Read-Through cache loads from DB on miss

Cache Invalidation

The two hard problems in computer science:

1. Cache invalidation
2. Naming things
3. Off-by-one errors

Use TTL-based expiration as the default. Event-driven invalidation for critical consistency.

Database Scaling

Vertical

Bigger machine. Simple. Limited.

Horizontal (Sharding)

• Hash-based consistent hashing for even distribution
• Range-based partition by date ranges or ID ranges
• Directory-based lookup table for shard mapping

Replication

• Primary-Replica writes to primary, reads from replicas
• Multi-Primary complex conflict resolution
• Trade-off: consistency vs. availability (CAP theorem)

Message Queues

When to use:

• Decouple services
• Handle traffic spikes (buffering)
• Async processing (emails, reports, exports)
• Event-driven architectures

Kafka high throughput, durable, ordered RabbitMQ flexible routing, simpler operations Redis Streams lightweight, good for simple queues

Key Principles

1. Design for failure everything will fail
2. Scale reads and writes independently
3. Prefer eventual consistency when possible
4. Measure before optimizing
5. Simple architectures > clever architectures