Understanding Caching: Concepts, Types, and Multi‑Level Cache Architecture

Traditional relational databases such as MySQL struggle with extreme traffic scenarios like flash sales or homepage spikes, and caching technologies provide an effective solution.

What is a cache? A cache stores frequently accessed hot data in memory, allowing subsequent requests to be served from memory instead of the database, thereby reducing database load and preventing crashes under high concurrency.

Why use a cache (and why Redis)? Caches improve both performance and concurrency. The first request loads data from the database into the cache; subsequent requests retrieve data from the fast in‑memory cache. Redis, for example, can handle up to 110,000 reads/s and 81,000 writes/s, dramatically increasing throughput.

Cache classifications include three main types: local cache, distributed cache, and multi‑level cache.

1. Local cache resides in the same process as the application, offering very fast read speed but limited storage capacity and no persistence across process restarts. It cannot be shared across multiple application instances, leading to consistency challenges that can be mitigated with Redis pub/sub for synchronization. Implementations often use key‑value structures or libraries such as Guava, Ehcache, or Caffeine.

2. Distributed cache runs as an independent service (e.g., Redis or Memcached) separate from the application, supporting large data volumes, persistence across restarts, and data consistency across clustered deployments. It enables read‑write separation and high availability through replication, though network latency makes it slower than local cache.

3. Multi‑level cache combines local and distributed caches to leverage the strengths of both. The typical request flow is: check the local (L1) cache; if miss, check the distributed (L2) cache; if still miss, query the database, then populate both caches. Implementations often use Guava or Caffeine for L1 and Redis for L2. In clustered environments, cache consistency can be maintained using Redis’s publish/subscribe mechanism.