When to use the Cache? #systemdesign #interviewpreparation #dsa #leetcode #faang

Cache is needed when you have a performance or cost problem caused by repeated expensive operations.

Use cache when:

1. Reads are frequent, data changes rarely — user profiles, config, product catalog. Fetching from DB every time wastes resources.

2. Computation is expensive — aggregations, ML inference, complex joins. Cache the result instead of recomputing.

3. External API calls are slow or rate-limited — third-party services, payment gateways. Cache responses to reduce latency and cost.

4. Database is a bottleneck — if DB CPU/connections are high due to repeated identical queries, a cache layer absorbs the read load.

5. Low tolerance for latency — sub-10ms response requirements that a DB (typically 10–100ms) can’t meet.

You probably don’t need cache when:

• Data changes on every request (cache hit rate would be near zero)

• Data must always be strongly consistent (e.g., bank balance)

• Your system is not yet at scale — premature caching adds complexity without benefit

The core trade-off:
Cache buys you speed and throughput at the cost of consistency (stale data risk) and complexity (invalidation logic, cache stampede, cold start).

A good rule of thumb:

measure first. If your DB query time or API latency is acceptable and you’re not hitting resource limits, you don’t need a cache yet. Add it when you have evidence of a bottleneck.

#systemdesign #backendengineering #tech #btech #faang

[ Cache, System design, Backend Engineering, Explore, Reels, Like for like, Support for new tech account ]

Видео When to use the Cache? #systemdesign #interviewpreparation #dsa #leetcode #faang канала Bytes Behind