Choose tools for scalable distributed systems

You are asked systems/design-concepts questions based on real product scenarios. For each scenario below: 1) Identify the *main requirements* (latency, throughput, durability, consistency, availability, cost, query patterns). 2) Choose an appropriate tool/architecture (DB/cache/queue/search/stream processing/etc.). 3) Justify trade-offs and list at least 2 risks/pitfalls. ## Scenarios ### A) Caching and read scalability You have a read-heavy service with hot keys and strict p99 latency. How would you use a cache (e.g., Redis) and what cache strategy would you pick (cache-aside, read-through, write-through, write-back)? How do you prevent cache stampede and handle invalidation? ### B) Concurrency control A user can place an order and you must avoid overselling inventory under high concurrency. - When is **optimistic locking** appropriate vs **pessimistic locking**? - If you use Redis for coordination, how would you implement a correct distributed lock? ### C) Asynchronous processing You need to process events (e.g., payments, notifications) reliably without slowing down the request path. When do you pick: - a message queue - a log-based streaming platform - a task scheduler / cron ### D) Storage selection Pick a primary storage solution and justify: - relational DB vs key-value store vs document store - when you would introduce sharding and what shard key you would choose ### E) Observability and performance What metrics and dashboards would you define to prove the system scales? How would you debug rising tail latency?