Deep dives into classic interview patterns + validated simulator results + whiteboard discussions. Every post is reproducible in the simulator.
Is service decomposition always the right answer? Same workload, monolith single server vs MSA chain — compare throughput, latency, and operational cost. Core whiteboard interview discussion.
Problems horizontal scale alone cannot fix: single DB bottleneck, cold cache spike, microservice chain latency, sticky sessions, connection pools, cascading failures. Each demonstrated in the simulator.
Hot-key contention. Concurrent updates on the same row trigger RDBMS row-lock serialization → backpressure → app threads exhausted → cascading failure. Four solutions (Redis INCR / sub-counter / Kafka serial / cell architecture).