Eviction Policy

When Redis or memcached runs out of memory, the behavior is defined.

Reject new writes. Drop the oldest entries. Drop the least recently used. LRU, LFU, TTL. The designer chooses a strategy. If no choice is made, a default kicks in. Either way, overflow is handled.

Human capacity works the same way.

Work, messages, health, relationships, hobbies, learning. You cannot keep all of it in memory. When it overflows, something must go. The problem is that humans have no eviction policy defined.

When overflow hits, you cannot predict what gets dropped. Important things do not necessarily survive. The urgent squats in place. The important but not pressing disappears quietly. By the time you notice, it is a cache miss.

With Redis, you review the config. Change maxmemory-policy. Humans have no such option. My memory is always overflowing.