Design a Thread-Safe In-Memory Cache (Deep Dive)

Problem

Apple's deep-dive round often goes below the language. Design a thread-safe, size-bounded in-memory cache and defend every decision around memory and concurrency.

Requirements

• get(key) / put(key, value, ttl)
• Bounded by entry count and/or total bytes
• TTL expiry
• Safe under concurrent readers and writers

What evaluators look for

• Lock granularity (single mutex vs sharded/striped locks vs read-write lock)
• Memory model: how you avoid retain cycles / leaks (ARC, weak refs)
• Eviction policy interaction with concurrency
• How you would expose metrics (hit rate, evictions)

Extension

Extend to a multi-process cache shared via mmap — what breaks?