A working sensor is useless to an application that cannot receive its data
An urban planning tool depends on an environmental data API. Sensors continue measuring, but API requests fail during peak demand. The gold standard is reliable delivery of appropriately qualified data, not merely an online endpoint.
Define service expectations
Availability, response time, data freshness and completeness should reflect the intended use.
Preserve clear data meaning
Units, timestamps, location identifiers and quality flags need consistent definitions.
Design for controlled failure
Appropriate caching, retries, rate limits and error messages can reduce disruption without concealing stale data.
Monitor real requests
Synthetic tests and operational metrics can reveal failures that server uptime alone misses.
Protect security and access
Authentication, permissions and appropriate usage controls support dependable operation.
Manage version changes
Schema changes should be communicated and tested so downstream applications do not fail silently.
Break-test: the API returns stale values with a success code
Can clients recognise the data age and avoid treating it as a current observation?
Connected knowledge
See Sensor Data Service Continuity, Sensor Network Resilience and NEA.