You rent construction equipment with sensors sending thousands of events per hour per device. You need consistent, time-ordered, atomic storage and retrieval per event. What should you do?

Cloud Bigtable is ideal for this scenario because it's a high-throughput, low-latency NoSQL database designed for large analytical and operational workloads, such as time-series data from IoT devices. Designing the row key with the event timestamp ensures consistent, time-ordered storage and efficient retrieval. Cloud Storage and Cloud Filestore are object and file storage services, respectively, not optimized for high-volume, real-time writes and atomic updates of individual events. Appending to files would be inefficient and lack the necessary atomic properties. Datastore (now Firestore in Datastore mode) offers atomic transactions within an entity group, but entity groups have a write rate limit of one transaction per second, which is insufficient for "thousands of events per hour per device.