Caching Basics

Caching stores a copy of frequently accessed data in a fast storage layer. When data is requested again, the cache serves it instead of recomputing or fetching from the slow database.

Analogy — Coffee shop:

Without cache: every latte requires roasting beans, grinding, and brewing (slow).
With cache: the barista pre-makes popular drinks and serves them instantly.

Where to Cache

flowchart LR
    Client["📱 Client Browser"] -->|"1. Browser Cache"| CDN["🌐 CDN Edge"]
    CDN -->|"2. CDN Cache"| Reverse["Reverse Proxy / LB"]
    Reverse -->|"3. Server Cache<br/>(Redis, Memcached)"| App["App Server"]
    App -->|"4. Database Cache<br/>(Buffer Pool)"| DB[("Database")]

    style Client fill:#7c3aed,color:#fff
    style CDN fill:#4f46e5,color:#fff
    style Reverse fill:#6366f1,color:#fff
    style App fill:#8b5cf6,color:#fff
    style DB fill:#dc2626,color:#fff

Cache Layer	Latency	Capacity	Example
Browser	~1 ms	Small	LocalStorage, HTTP cache headers
CDN	~10-50 ms	Medium	CloudFlare, Fastly, CloudFront
Server Cache	~1-5 ms	Medium	Redis, Memcached
Database Cache	~0.1 ms	Large	InnoDB buffer pool

What to Cache

Data Type	Cache?	Why
Static assets (images, CSS, JS)	✅ Yes	Never changes, serve from CDN
User session data	✅ Yes	Read on every request, fast access
Database query results	✅ Yes	Expensive queries, slow to recompute
User feed (personalized)	✅ Maybe	Cache for a few seconds, then refresh
Real-time stock prices	❌ No	Changes too fast, stale data is bad
Passwords / sensitive data	❌ No	Security risk

Cache Performance Impact

Scenario	Latency
Cache hit (Redis, in-memory)	~1-5 ms
Cache miss (query database)	~10-100 ms
Cache hit (CDN, edge)	~10-50 ms
No cache (full origin)	~100-500 ms

A cache hit is 10-100× faster than a database query.

Trade-offs

Cache adds complexity (another system to manage).
Stale data: cache might serve outdated content.
Cache breakdown: if the cache goes down and all requests hit the database at once (thundering herd), the database can crash.

In Simple Words

Cache = storing copies of slow-to-get data in a fast place (memory).
Cache data that’s read often and changes rarely.
A cache hit takes milliseconds. A cache miss costs a database query (tens of milliseconds).