부산대학교 도서관

An IoT system of water management for automated irrigation in agriculture can use sensors to obtain metrics such as soil moisture, soil temperature, soil pH, air humidity, and air temperature to make more precise decisions in irrigation. Fog computing can store and analyze data during cloud disconnection and providing system availability due to Internet disconnections, expected in a farm scenario. However, fog nodes have resource constraints. Also, depending on the crop type or crop growth stage, not all collected data are relevant for irrigation. In some cases, the IoT system uses multi-depth sensors to collect data. Still, depending on the plant root size, some deeper measurements collect irrelevant data and increase the demand for memory in IoT devices. This paper evaluates a data value mechanism to extract relevant data in a fog-based IoT system for smart irrigation. We consider the system with a workload of 500 data packets per minute during cloud network availability and unavailability. We create synthetic input data using three soil moistures for different depths and permute data value configurations to consider one, two, or three depths values as relevant data in the experiment. Our results show a data size reduction of 23.15% or 47.60%, depending on the crop growth stage. We also perceive a statistical tie for packet delay and batch transfer time metrics with all configurations.