Vehicle sensors send events from many small regional data centers to central processing over unreliable leased lines with unpredictable latency. What is the most cost-effective way to improve event delivery reliability?

Cloud Pub/Sub is a globally distributed messaging service designed for high availability and durability, making it ideal for handling unreliable network conditions. By having devices publish directly to Pub/Sub, messages are reliably ingested and stored, even if the connection to the central processing is intermittent. This decouples the data producers from consumers, ensuring data isn't lost during network outages. Deploying Kafka clusters adds operational overhead and cost for managing infrastructure in each data center. Cloud Interconnect improves network performance but doesn't inherently solve the problem of unreliable leased lines causing data loss if the connection drops. A Cloud Dataflow pipeline is for processing data, not for reliably ingesting it from edge devices over unreliable networks.