Cache eviction strategies (TTL, LRU)

When the cache is full or an entry is stale, something must be removed — that’s eviction. Common strategies: TTL (time-to-live): remove after a fixed time; LRU (least recently used): remove the item that hasn’t been used for the longest time.

TTL: time-based expiry

flowchart LR Set[Set key, TTL=60s] --> Store[Store + expiry time] Store --> Check{Read request} Check -->|Now < expiry| Hit[Cache hit] Check -->|Now >= expiry| Miss[Evict, cache miss]

LRU: evict least recently used

flowchart TB subgraph LRU["LRU list (order = last access)"] A[Most recent] --> B --> C --> D[Least recent] end Full[Cache full] --> Evict[Evict D] Evict --> New[Add new entry at head]

Strategy	When evict	Use case
TTL	After fixed duration	Stale-safe data (e.g. 5 min for profile)
LRU	When full; drop least recently used	Bounded memory; keep “hot” data
LFU	When full; drop least frequently used	Keep most popular items
FIFO	When full; drop oldest inserted	Simple, no access tracking

Combine TTL (so data doesn’t stay forever wrong) with LRU (so under memory pressure you drop the coldest entries). Many caches (e.g. Redis) support both.