Redis Memory Management: Eviction and Expiration Strategies Explained

Eviction Strategy Principles

Redis uses the maxmemory setting to limit memory usage; when the limit is reached, it applies an eviction policy defined by maxmemory-policy to free space for new data.

maxmemory Configuration

# client command to view and set maxmemory
127.0.0.1:6379> config get maxmemory
"maxmemory"
"0"
127.0.0.1:6379> config set maxmemory 100mb
OK

If maxmemory=0 , no limit is enforced; on 32‑bit Windows the default usable space is about 3 GB and the policy defaults to noeviction , causing write errors when memory is exhausted.

Eviction Policy Types

Redis defines eight eviction policies (volatile‑lru, volatile‑ttl, volatile‑random, allkeys‑lru, allkeys‑random, noeviction, and LFU variants introduced in Redis 4.0).

# command‑line configuration of eviction policy
CONFIG GET maxmemory-policy
"noeviction"
CONFIG SET maxmemory-policy volatile-lru
OK

freeMemoryIfNeeded Logic (excerpt)

int freeMemoryIfNeeded(void) {
  size_t mem_used, mem_tofree, mem_freed;
  // Compute used memory, subtract slave output buffers and AOF buffers
  // If memory exceeds maxmemory, select keys according to the configured policy
  // and delete them, updating statistics and notifying replicas as needed.
}

Expiration Strategy Principles

Keys with an expiration are stored in the internal expires dictionary; Redis checks expiration on access and also via periodic active expiration and lazy deletion.

# expiration commands
expire key seconds
pexpire key milliseconds
expireat key timestamp
pexpireat key millisecond_timestamp

Active expiration runs at a configurable frequency (default 10 times per second) scanning a random subset of keys, while lazy expiration checks a key’s TTL during each read/write operation.

Choosing an Eviction Policy

If hot and cold data coexist, prefer allkeys‑lfu with lfu-decay-time to avoid cache pollution.

For uniformly accessed data, allkeys‑random provides simple random eviction.

To keep certain data longer, use volatile‑lfu (Redis 4.0+) or volatile‑lru (Redis 3.0‑).

When ordering by expiration time is required, choose volatile‑ttl .

Combining periodic active expiration with lazy deletion offers a balanced approach that efficiently reclaims memory while preserving data consistency.