Technical detail · Sensing as a Service
Under the Hood: sensor-cloud composition and caching
Three actors sit on either side of the boundary. The sensor-owner purchases and deploys the physical hardware and lends it to the service. The sensor-cloud service provider (SCSP) takes on redeployment, maintenance, and overhead, and manages the virtualization layer between owners and consumers. The end-user consumes the service through an application and is not expected to know which physical sensor answered the request. Naming these three roles explicitly is what the “who controls the physical sensor” and “who exposes the service interface” evidence questions above are asking for.
The virtualization layer composes physical sensors into virtual sensors at two granularities: CoV-I groups homogeneous physical nodes within the same geographic region into one virtual sensor, and CoV-II spans heterogeneous nodes across multiple regions into a virtual sensor group. Composition granularity is itself review evidence — a consumer decision that assumes a single-region reading should not silently receive a multi-region composite without knowing it.
The layer also runs an internal and external caching mechanism motivated by a specific observation: physical conditions often change slowly enough that constant re-sensing wastes energy for no gain. An internal cache answers end-user requests directly when it can, and only pulls a re-cache from an external cache — which itself refreshes from the physical layer on an interval — when the cached value is no longer good enough. This is a concrete version of the quality-evidence and change questions above: the consumer’s freshness requirement decides how far back through the cache the request has to travel.