Redis on EC2: Instance Selection Guide¶

Redis is a high-performance in-memory data structure store used as a database, cache, and message broker. Choosing the right EC2 instance is crucial for optimal performance and cost efficiency.

Instance Requirements for Redis¶

Memory Considerations¶

Redis stores all data in memory, making memory the primary consideration:

• Memory-to-data ratio: Plan for 2-3x your dataset size
• Memory speed: Faster memory improves performance
• NUMA considerations: Important for large instances

CPU Requirements¶

Redis is primarily single-threaded, but modern versions can utilize multiple cores:

• Single-core performance: More important than core count
• ARM vs x86: Graviton processors offer excellent price/performance

Recommended Instance Types¶

Small to Medium Deployments (< 10GB)¶

r7g.large - Best overall choice - 2 vCPUs (ARM Graviton3) - 16 GB memory
- Up to 12.5 Gbps network - Excellent price/performance ratio

r6i.large - x86 alternative - 2 vCPUs (Intel Xeon) - 16 GB memory - Up to 12.5 Gbps network - Better single-thread performance

Large Deployments (10-100GB)¶

r7g.xlarge - ARM performance - 4 vCPUs (ARM Graviton3) - 32 GB memory - Up to 12.5 Gbps network

r6i.xlarge - Intel performance
- 4 vCPUs (Intel Xeon) - 32 GB memory - Up to 12.5 Gbps network

Performance Comparison¶

*US East prices, subject to change

Configuration Best Practices¶

Memory Management¶

# Configure Redis memory limit (leave 25% for OS)
maxmemory 12gb
maxmemory-policy allkeys-lru

Persistence Configuration¶

# For cache workloads - disable persistence
save ""

# For persistent data - configure appropriately  
save 900 1
save 300 10
save 60 10000

Network Optimization¶

# Increase network buffers for high throughput
tcp-backlog 511
tcp-keepalive 300

Monitoring and Scaling¶

Key Metrics to Monitor¶

• Memory usage and fragmentation
• CPU utilization
• Network throughput
• Keyspace hit rate
• Connected clients

Scaling Strategies¶

1. Vertical scaling: Upgrade to larger instance
2. Read replicas: Add read-only replicas
3. Clustering: Use Redis Cluster for horizontal scaling
4. Sharding: Application-level data partitioning

Cost Optimization¶

Reserved Instance Savings¶

• Standard RIs: Up to 60% savings
• Convertible RIs: Up to 45% savings with flexibility

Spot Instance Considerations¶

Generally not recommended for Redis due to: - Data loss risk on interruption - Warm-up time after restart - Use only for development environments

Conclusion¶

For most Redis deployments, r7g.large offers the best price/performance ratio. Scale up to r7g.xlarge or r7g.2xlarge as memory requirements grow. Always monitor performance and adjust instance types based on actual usage patterns.