While most systems are read-heavy, some of the most challenging system design interviews involve massive write loads: GPS pings from drivers, clickstream events, server logs, metrics, trades, orders, and IoT readings.

High-write systems are hard because every accepted write forces you to answer a few concrete questions: when is it safe to acknowledge, where is the write ordered, what happens if a node fails, and how do you prevent one partition from becoming the bottleneck?

Reads can often be served from caches or replicas. Writes can be buffered, batched, partitioned, and sometimes acknowledged before final persistence, but they cannot be ignored. If the product says a write succeeded, the system needs either a durable path for it or an explicit decision that some loss is acceptable.