5.2 Redis Memory Management

1. Breakdown of Memory Consumption

1.1. used_memory

• used_memory = Redis itself + object memory + buffer memory

1.1.1. Redis Itself

The Redis process, generally negligible.

1.1.2. Object Memory

• sizeof(key) + sizeof(value), which takes the largest share.
• Solutions
  • Use serialization to reduce object size.
  • Encoding optimization.

1.1.3. Buffer Memory

buffer memory = client buffer + replication backlog buffer + AOF buffer

1.2. Memory Fragmentation

• used_memory_rss - used_memory = memory fragmentation
  • Modern memory allocators use fixed-range memory blocks and round allocations upward.
• Solution
  • Restarting can reorganize it.

2. How to Inspect Memory Consumption

• Use info memory to view used_memory_rss, used_memory, and mem_fragmentation_ratio.
  • mem_fragmentation_ratio = used_memory_rss / used_memory
  • If mem_fragmentation_ratio > 1, memory fragmentation exists.
  • If mem_fragmentation_ratio < 1, Linux has swapped Redis to disk.

3. How to Manage Memory

3.1. Memory Eviction Strategy

• Set max_memory + a memory eviction strategy.
• Redis Memory Eviction Strategy

3.2. Key Expiration

• Redis Key Expiration Strategy

3.3. Memory Optimization

• Redis Memory Optimization