Remote Monitoring Systems: Sensors, Alerts and Dashboards
Remote monitoring is one of the most common IoT applications: temperatures in cold stores, levels in tanks, vibration on pumps, power in remote cabinets. The parts are familiar, a sensor, a connection, a database and a dashboard, but systems succeed or fail on details: whether readings can be trusted, whether silence is noticed, and whether alerts are rare enough that people still act on them.
At a glance
- Heartbeat
- A regular message proving the device is alive, so silence itself can raise an alert.
- Hysteresis
- Separate trigger and clear levels stop an alarm chattering around a threshold.
- Delay
- Requiring a condition to persist before alerting filters out brief spikes.
- UTC
- Storing timestamps in UTC avoids confusion across time zones and clock changes.
The problem
The common failure is not a missed reading but a missed absence. A device that loses power or signal stops sending, and a system that alerts only on bad values reports nothing, so a failed freezer and a failed sensor look the same: quiet.
The opposite failure is alert fatigue. Thresholds set without hysteresis or delay generate repeated notifications for marginal conditions, people learn to ignore them, and the one real alarm is ignored with the rest.
Treat silence as a reading. A system that cannot tell a quiet site from a dead sensor is not monitoring.
Designing alerts people act on
Each alert needs a clear condition, a persistence time, a separate clear condition, an owner and an escalation path if nobody acknowledges it. Missing data, low battery and sensor faults deserve alerts of their own, routed to whoever maintains the devices.
Reporting interval sets both responsiveness and running cost; the first figure shows messages per device per day at common intervals. Many products report on a slow interval and send immediately when a reading crosses a threshold, keeping traffic low without delaying alarms.
What good looks like
For each measured value, define the normal range, alarm conditions, persistence and clear levels, and who must be told.
Add heartbeat and data-quality monitoring for every device, with alerts that reach the people responsible for maintaining them.
Review alert history regularly and tune any alert that fires often without leading to action.
| messages | |
|---|---|
| Every 10 s | 8640 |
| Every minute | 1440 |
| Every 5 minutes | 288 |
| Every 15 minutes | 96 |
| Every hour | 24 |
From reading to response
- Measure: Reading with timestamp
- Check: Range, persistence, data quality
- Alert: Notify the owner
- Escalate: If not acknowledged
- Clear: When the condition resolves
What it changes
Trusted Alerts
Notifications are rare and meaningful, so people respond to them.
No Silent Failures
Dead devices and lost connections are detected as quickly as bad readings.
Controlled Costs
Reporting intervals match the response time the application needs.
Useful Dashboards
Views answer operational questions rather than plotting every value.
Where to go next
Calibrate sensors and record calibration dates. A well-designed alert on an inaccurate reading is still wrong.
Plan retention and downsampling: keep raw data for a defined period and longer-term summaries for trends.
Problems this prevents
- Dead sensors mistaken for quiet sites
- Alarms chattering around a threshold
- Alerts with no owner or escalation
- Timestamps confused by time zones
Discuss your project
Email support@aavico.com or use Get in Touch when the website finishes loading.