In-memory cache

The in-memory cache stores verification results inside each Ory Talos process. It uses ristretto with TinyLFU admission and sampled LFU eviction.

Configuration

cache:
  type: "memory"
  ttl: "5m"
  memory:
    max_size: 104857600 # Maximum cache size in bytes (100 MiB)
    num_counters: 10000 # Counters for TinyLFU frequency estimation

cache.type and the entire cache.memory.* block are immutable. Changes require a server restart to take effect.

Characteristics

• Per-process: Each Ory Talos instance has its own cache. It isn't shared across instances.
• Bounded by max_size: When the cache reaches max_size bytes, TinyLFU admission and sampled LFU eviction decide which entries to keep.
• TTL-based expiry: Entries expire after the configured ttl.
• No external dependencies: The cache runs in the Talos process, with no network round trips.

When to use

Use the in-memory cache for single-node deployments, or when each instance handles enough traffic to benefit from local caching on its own. For multi-instance deployments, use Redis to share a cache across all instances.