<![CDATA[Many hydroponic Internet-of-Things prototypes display sensor readings but provide limited evidence that each record remains complete and traceable across the full local data path. This paper evaluates a self-hosted edge-IoT pipeline in which an ESP32 publishes pH and electrical conductivity (EC) measurements through Mosquitto MQTT to Telegraf, InfluxDB, and Grafana on a Raspberry Pi 4. Four traceable datasets were used: 80 calibration observations, 2,881 paired pH–EC records over approximately 48 h, three 30-min high-rate storage trials, and 43 broker-to-dashboard matched observations. The pH mean absolute percentage error (MAPE) was 0.775% and 0.386% at pH 4.00 and 7.00; EC MAPE was 2.173% and 0.986% at 1,413 and 12,880 μS cm−1. The 48-h export contained all 2,881 expected one-minute records. The high-rate trials stored 1,018 of 1,020 expected records, corresponding to 99.80% completeness. All 43 broker values were matched in Grafana. Their absolute timestamp-difference proxy had a mean of 345.91 ms and a 95th percentile of 763.6 ms; it is not claimed as true end-to-end latency because Grafana timestamps were exported at one-second resolution. The resultsdemonstrate a compact and auditable local monitoring pipeline while identifying the instrumentation still requiredfor defensible delay and loss attributio.]]>
Copyrights © 2026
