What cache invalidation strategy would you choose?

The choice of cache invalidation strategy depends on data freshness requirements and system load. Main approaches:

• Time-to-Live (TTL) — cache is automatically considered stale after a specified time. Good for data that changes infrequently.
• Event-based invalidation — cache is explicitly cleared or updated when data changes in the source. Requires a notification mechanism.
• Write-through / Write-back — cache is updated synchronously or asynchronously during data writes.
• Cache-aside — the application checks the cache itself and, if data is missing or stale, queries the source and updates the cache.

For complex systems, TTL is often combined with event-based invalidation to balance data freshness and performance.

For example, in a distributed system, a publish-subscribe mechanism (e.g., via a message broker) can be used to notify all nodes to invalidate cache upon data changes.