Cache invalidation strategies: cách xử lý data freshness trong Redis?

Cache invalidation là bài toán khó nổi tiếng (Phil Karlton: 'there are only two hard things in Computer Science: cache invalidation and naming things').

Các chiến lược: TTL-based expiration: đơn giản nhất — set TTL phù hợp, data stale tối đa TTL seconds.
Phù hợp cho data có freshness requirement rõ ràng. Event-driven invalidation: khi data trong DB thay đổi, publish event → subscriber DEL cache key tương ứng ngay.
Dùng Debezium CDC hoặc domain events.
Chính xác hơn TTL nhưng phức tạp hơn. Write-through invalidation: khi write DB, đồng thời DELETE hoặc SET cache mới. Versioned keys: thêm version vào cache key (product:123:v5), khi update thì tăng version — key cũ tự expire theo TTL; không cần explicit invalidation. Cache tags (advanced): nhóm nhiều cache key theo tag (tag:category:electronics), khi category thay đổi thì invalidate tất cả key có tag đó — Redis không native support, cần implement bằng Set lưu list key per tag.
Thực tế: kết hợp short TTL (30-60s) + event invalidation để balance freshness và complexity.

Cache invalidation is a famously hard problem (Phil Karlton: 'there are only two hard things in Computer Science: cache invalidation and naming things').

Strategies: TTL-based expiration: the simplest — set an appropriate TTL, data is stale for at most TTL seconds.
Works well when freshness requirements are clear. Event-driven invalidation: when data changes in the DB, publish an event → a subscriber DELetes the corresponding cache key immediately.
Use Debezium CDC or domain events.
More accurate than TTL but more complex. Write-through invalidation: on every DB write, simultaneously DELETE or SET the new cache value. Versioned keys: embed a version in the cache key (product:123:v5); on update, increment the version — old keys expire naturally by TTL; no explicit invalidation needed. Cache tags (advanced): group cache keys under a tag (tag:category:electronics); when the category changes, invalidate all keys with that tag — Redis doesn't support this natively, so implement it using a Set that stores the list of keys per tag.
In practice: combine short TTLs (30-60s) with event invalidation to balance freshness and complexity.

Xem toàn bộ Redis cùng filter theo level & chủ đề con.

Mở danh sách Redis

Cache invalidation là bài toán khó nổi tiếng (Phil Karlton: 'there are only two hard things in Computer Science: cache invalidation and naming things').

Các chiến lược: TTL-based expiration: đơn giản nhất — set TTL phù hợp, data stale tối đa TTL seconds.
Phù hợp cho data có freshness requirement rõ ràng. Event-driven invalidation: khi data trong DB thay đổi, publish event → subscriber DEL cache key tương ứng ngay.
Dùng Debezium CDC hoặc domain events.
Chính xác hơn TTL nhưng phức tạp hơn. Write-through invalidation: khi write DB, đồng thời DELETE hoặc SET cache mới. Versioned keys: thêm version vào cache key (product:123:v5), khi update thì tăng version — key cũ tự expire theo TTL; không cần explicit invalidation. Cache tags (advanced): nhóm nhiều cache key theo tag (tag:category:electronics), khi category thay đổi thì invalidate tất cả key có tag đó — Redis không native support, cần implement bằng Set lưu list key per tag.
Thực tế: kết hợp short TTL (30-60s) + event invalidation để balance freshness và complexity.

Cache invalidation is a famously hard problem (Phil Karlton: 'there are only two hard things in Computer Science: cache invalidation and naming things').

Strategies: TTL-based expiration: the simplest — set an appropriate TTL, data is stale for at most TTL seconds.
Works well when freshness requirements are clear. Event-driven invalidation: when data changes in the DB, publish an event → a subscriber DELetes the corresponding cache key immediately.
Use Debezium CDC or domain events.
More accurate than TTL but more complex. Write-through invalidation: on every DB write, simultaneously DELETE or SET the new cache value. Versioned keys: embed a version in the cache key (product:123:v5); on update, increment the version — old keys expire naturally by TTL; no explicit invalidation needed. Cache tags (advanced): group cache keys under a tag (tag:category:electronics); when the category changes, invalidate all keys with that tag — Redis doesn't support this natively, so implement it using a Set that stores the list of keys per tag.
In practice: combine short TTLs (30-60s) with event invalidation to balance freshness and complexity.