A globally distributed auction app receives identical near-simultaneous bids processed by different servers. Each bid has item, amount, user, timestamp. You must collate bids in real time and determine which user bid first. What should you do?

The correct approach leverages Cloud Pub/Sub for reliable, real-time ingestion of bid events from all servers, ensuring no bids are lost. Dataflow then processes these events in a streaming fashion. Dataflow's ability to handle out-of-order events and process them based on event time (the bid's timestamp) is crucial for accurately determining the first bid, even if network latency causes processing delays. This architecture is scalable and fault-tolerant, suitable for a globally distributed application. Writing to a shared file and using Hadoop would introduce significant latency and complexity for real-time processing. Using Cloud SQL or MySQL instances per server would struggle with the high write volume and real-time consistency requirements for determining the absolute first bid across distributed servers.