Distributed caching (Redis, Memcached) in HLD - System Design Exercise

Design: Distributed Caching System

Design the caching layer architecture using Redis or Memcached for a web application backend. Out of scope: database design, client application logic, and detailed security mechanisms.

Functional Requirements

FR1: Cache frequently accessed data to reduce database load

FR2: Support high read throughput with low latency (p99 < 10ms)

FR3: Handle 100,000 concurrent cache requests

FR4: Provide cache consistency with eventual consistency model

FR5: Support cache invalidation and expiration policies

FR6: Ensure high availability with 99.9% uptime

FR7: Allow horizontal scaling of cache nodes

Non-Functional Requirements

NFR1: Cache size limited to 100GB per node

NFR2: Network latency between cache nodes and clients < 5ms

NFR3: Data loss acceptable for short periods (cache is a performance layer)

NFR4: Cache miss penalty should not exceed 50ms additional latency

Think Before You Design

Questions to Ask

❓ Question 1

❓ Question 2

❓ Question 3

❓ Question 4

❓ Question 5

❓ Question 6

❓ Question 7

Key Components

Cache nodes (Redis or Memcached instances)

Cache clients integrated with application servers

Cache coordinator or proxy (optional)

Cache invalidation mechanism

Monitoring and alerting system

Load balancer or DNS for cache node discovery

Design Patterns

Cache-aside pattern

Write-through and write-back caching

Consistent hashing for cache sharding

Replication for high availability

Cache warming and preloading

Cache invalidation strategies

Reference Architecture

Application Servers

↓

Cache Clients

↓

Cache Node 1

↓

Persistent DB

Components

Cache Nodes

Redis or Memcached

Store cached data in memory for fast access with support for sharding and replication

Cache Clients

Language-specific Redis/Memcached clients

Interface between application servers and cache nodes to read/write cached data

Application Servers

Any backend framework

Serve user requests, query cache first, fallback to database on cache miss

Persistent Database

Relational or NoSQL DB

Source of truth for data, updated on writes

Monitoring & Alerting

Prometheus, Grafana, or similar

Track cache hit rates, latency, node health, and trigger alerts

Request Flow

1. Client sends request to Application Server

2. Application Server queries Cache Client for data using cache key

3. Cache Client checks Distributed Cache Cluster

4. If cache hit, data returned immediately to Application Server

5. If cache miss, Application Server queries Persistent Database

6. Application Server updates cache with fresh data (cache-aside pattern)

7. Application Server returns data to Client

8. Cache nodes replicate data asynchronously for availability

9. Cache entries expire or are evicted based on TTL or LRU policies

Database Schema

Not applicable - caching layer stores key-value pairs with keys representing data identifiers and values as serialized data objects. No fixed schema required.

Scaling Discussion

Bottlenecks

Single cache node memory limits causing cache misses

Network bandwidth saturation between application and cache nodes

Cache node failure causing data unavailability

Cache stampede on popular keys causing database overload

Inconsistent cache data due to delayed invalidation

Solutions

Use consistent hashing to shard cache keys across multiple nodes to scale memory horizontally

Deploy cache nodes in the same data center or availability zone to reduce network latency

Implement replication and failover mechanisms for cache nodes to ensure high availability

Use request coalescing or locking to prevent cache stampede on popular keys

Adopt cache invalidation strategies like write-through or event-driven invalidation to maintain freshness

Interview Tips

Time: Spend 10 minutes clarifying requirements and constraints, 20 minutes designing architecture and data flow, 10 minutes discussing scaling and trade-offs, 5 minutes summarizing and answering questions.

Explain cache-aside pattern and why it fits many use cases

Discuss trade-offs between Redis and Memcached (persistence, data structures)

Highlight importance of cache invalidation and consistency models

Describe how consistent hashing helps scale cache horizontally

Mention monitoring and alerting to maintain cache health

Address failure scenarios and recovery strategies