Redis Architecture Design Guide — Standalone to Cluster

Three Redis Architectures

Type	Standalone	Sentinel	Cluster
Nodes	1	3+	6+ (minimum)
High Availability	No	Yes (Failover)	Yes (Failover)
Horizontal Scaling	No	No	Yes (Sharding)
Complexity	Low	Medium	High
Suitable Scale	Dev / Small	Small–Medium	Large

1. Standalone

Single node. Any failure causes a full service outage.

# docker-compose.yml
services:
  redis:
    image: redis:7-alpine
    ports:
      - "6379:6379"
    command: redis-server --maxmemory 256mb --maxmemory-policy allkeys-lru
    volumes:
      - redis-data:/data

When to use: Development environments, small services where sessions or simple caching are the only use case.

2. Sentinel (Recommended for Small–Medium Scale)

1 Master + N Replicas + 3 Sentinel nodes.
On Master failure, Sentinel automatically promotes a Replica to Master.

[Client] → [3 Sentinel nodes] → [Master]
                                   ↓ Replica

# sentinel.conf
sentinel monitor mymaster 192.168.1.10 6379 2
sentinel down-after-milliseconds mymaster 5000
sentinel failover-timeout mymaster 60000
sentinel parallel-syncs mymaster 1

Spring Boot Connection:

spring:
  data:
    redis:
      sentinel:
        master: mymaster
        nodes:
          - 192.168.1.1:26379
          - 192.168.1.2:26379
          - 192.168.1.3:26379
      password: your-password

3. Cluster (Large Scale / Horizontal Scaling)

Data distributed across 16384 hash slots. Minimum: 3 Masters + 3 Replicas = 6 nodes.

Slots 0–5460    → Master1 (+ Replica1)
Slots 5461–10922  → Master2 (+ Replica2)
Slots 10923–16383 → Master3 (+ Replica3)

# Create cluster
redis-cli --cluster create \
  192.168.1.1:6379 192.168.1.2:6379 192.168.1.3:6379 \
  192.168.1.4:6379 192.168.1.5:6379 192.168.1.6:6379 \
  --cluster-replicas 1

Note: Multi-key commands (MGET, MSET, etc.) only work when all keys are on the same slot.
Solution: Use Hash Tags → {user:1}:profile, {user:1}:session

Memory Eviction Policy

# maxmemory-policy options
allkeys-lru      # Evict least recently used keys (best for caching)
volatile-lru     # Evict only keys with TTL, by LRU
allkeys-lfu      # Redis 4.0+ LFU algorithm
noeviction       # No eviction — return write error (session store)

Data Persistence

# RDB (snapshot) — performance-first
save 900 1      # Save if 1+ key changed in 900 seconds
save 300 10
save 60 10000

# AOF (write-ahead log) — durability-first
appendonly yes
appendfsync everysec  # fsync every second (balanced)

Essential Monitoring Metrics

# Real-time stats
redis-cli info stats | grep -E "instantaneous|ops_per_sec|rejected"

# Memory usage
redis-cli info memory | grep -E "used_memory_human|maxmemory_human"

# Slow log (commands over 10ms)
redis-cli slowlog get 10

Architecture Selection Guide

• DAU < 10K: Standalone
• DAU 10K–1M: Sentinel (Master 1 + Replica 2)
• DAU > 1M or data > 100GB: Cluster
• Session store: Sentinel + noeviction
• Cache layer: Standalone or Cluster + allkeys-lru