Data loss implications

There are many scenarios we could evaluate to find the optimal delivery guarantee for a given process. As an example, each time a mechanical barrier counter increments, this is sent via an event to the maintenance tracking service. If each event simply contains a barrier ID, then we have the decision to make in the accuracy of the counter driving the maintenance schedule.

At-most-once delivery will ensure we do not overcount the use of each barrier due to system failures that would otherwise result in duplicate events. However, it also means we might undercount for the same reasons. This could result in late maintenance, meaning a higher probability of barrier failure.

At-least-once delivery will ensure we do not miss any counts in the use of a barrier. However, it also means we might overcount through duplicate (retry) events. This could result in early maintenance, meaning a higher cost to operate. It will also impact system scalability and throughput...