Multi-Layer Caching Strategy with Redis and CDN

Design a multi-layer caching strategy for [PROJECT NAME], a [APPLICATION TYPE] with [NUMBER] daily requests and traffic spikes of [NUMBER]x normal volume.\\\\\\\\n\\\\\\\\n**Current Stack:**\\\\\\\\n- Backend: [NODE.JS/PYTHON] + [FRAMEWORK]\\\\\\\\n- Database: [POSTGRESQL/MYSQL/MONGODB]\\\\\\\\n- Hosting: [AWS/GCP/VPS]\\\\\\\\n- CDN: [CLOUDFLARE/CLOUDFRONT/NONE]\\\\\\\\n\\\\\\\\n**Layer 1 — CDN/Edge Cache:**\\\\\\\\n- What to cache: static assets, SSG pages, immutable API responses\\\\\\\\n- Cache-Control headers by resource type\\\\\\\\n- Stale-while-revalidate for semi-dynamic content\\\\\\\\n- Purge strategy (by URL, by tag, full)\\\\\\\\n- Edge functions for personalization without origin requests\\\\\\\\n\\\\\\\\n**Layer 2 — Application Cache (Redis):**\\\\\\\\n- Architecture: Redis standalone vs. cluster vs. Redis Sentinel\\\\\\\\n- Cache patterns:\\\\\\\\n  - Cache-aside (lazy loading): read cache → miss → query DB → populate cache\\\\\\\\n  - Write-through: write to DB and cache simultaneously\\\\\\\\n  - Write-behind: write to cache → async to DB (for counters/analytics)\\\\\\\\n- TTL strategy by data type:\\\\\\\\n  - Session data: [DURATION]\\\\\\\\n  - Catalog data: [DURATION]\\\\\\\\n  - User data: [DURATION]\\\\\\\\n  - Computed data (rankings, stats): [DURATION]\\\\\\\\n- Serialization: JSON vs. MessagePack vs. Protocol Buffers\\\\\\\\n- Key namespacing: \\\\\\\\\\\\\\\\\\\\\\\`app:domain:id:field\\\\\\\\\\\\\\\\\\\\\\\`\\\\\\\\n- Eviction policy: allkeys-lfu vs. volatile-lru\\\\\\\\n\\\\\\\\n**Layer 3 — Query/ORM Cache:**\\\\\\\\n- Prisma/TypeORM/Drizzle: frequent query caching\\\\\\\\n- Materialized views for aggregations\\\\\\\\n- Connection pooling (PgBouncer)\\\\\\\\n\\\\\\\\n**Cache Invalidation (the hard part):**\\\\\\\\n- Event-based invalidation (pub/sub)\\\\\\\\n- Tag-based invalidation (e.g., invalidate everything for \\\\\\\\\\\\\\\\\\\\\\\`product:123\\\\\\\\\\\\\\\\\\\\\\\`)\\\\\\\\n- Versioned keys for zero-downtime deployments\\\\\\\\n- Race conditions: thundering herd problem → cache stampede lock\\\\\\\\n- Debugging: how to know if serving from cache or origin\\\\\\\\n\\\\\\\\n**Metrics and Monitoring:**\\\\\\\\n- Target cache hit ratio: > [%]\\\\\\\\n- Dashboards: hit rate, miss rate, eviction rate, latency\\\\\\\\n- Alerts: hit ratio < 80%, Redis latency > 5ms\\\\\\\\n- Estimated infrastructure cost savings after implementation\\\\\\\\n\\\\\\\\nInclude code for the cache middleware and helper functions in [LANGUAGE].