Connections, devices, and metrics – Observis

Observis organizes incoming monitoring data into three levels: connections, devices, and metrics. Learning them once explains most of the interface, because the same three words appear when you build a dashboard card, define a calculation, and configure an alert.

Level	What it is
Connection	The configured link that polls one data source on your behalf.
Device	An entry that reports readings through a connection, identified by an EUI.
Metric	One named measurement on a device, such as Displacement or Strain.

Connection

A connection is the configured link between your project and one monitoring data source. It holds the connection type, the credentials and endpoint for that source, and the polling schedule. It belongs to exactly one project.

When you create a connection you are setting up polling on your behalf: from then on, Observis contacts the source on the configured schedule and the resulting readings flow into your project without further action.

A connection uses a single connection type, so all devices on one connection come from the same vendor. Monitoring a site with hardware from two vendors means two connections.

One connection usually delivers many devices. On a large site, a single datalogger connection can account for dozens.

Note: Readings keep the connection they arrived on permanently. If a connection is later reconfigured or removed, the historical data it already collected stays in the project.

Device

A device is an entry that reports readings through a connection. Each one is identified by an EUI, a unique identifier assigned to the hardware at manufacture.

The important thing to know is that a device does not always correspond to one physical box. How readings are divided into devices depends on the connection type, and three patterns all occur in practice:

Pattern	What you see	Example
One device, many metrics	A single device carrying every reading from the unit	A vibration monitor reporting peak particle velocity, frequency, acceleration, and battery as metrics on one device
One device per channel	The physical node appears as several devices, one per wired-in sensor	A six-channel wireless node appearing as six devices that share a name prefix
One device per metric group	Readings from one logger split across devices by category	A datalogger whose battery voltage and internal temperature arrive as two separate devices

This is why a project can list more devices than you have instruments installed, and why two connection types can present the same physical setup differently. Neither is a fault.

You do not create devices by hand. As data arrives through a connection, Observis registers the devices the source reports. A new device appearing usually means the source started reporting something new.

Each device has a type, shown in the Type column of Pipeline Manager's table view — saa_displacement, dewesoft_channel, tiltmeter and so on. These come from the source system, which is why they read like identifiers rather than labels. The type tells you what kind of readings to expect from that device.

Note: Devices can register before any readings are stored, so a device with no data is not necessarily broken. If devices appear but stay empty, look at the connection rather than the instrument.

Metric

A metric is one named measurement on a device, with a unit. Examples include Displacement, Strain, Pressure, Temperature external, Thermistor resistance, and VW frequency A.

Which metrics a device offers is reported by the source, not configured in Observis. That list is why a device selector is always followed by a metric selector.

Observis has no separate concept of a sensor. A physical sensor appears either as a metric on a device or as a device of its own, depending on the connection type. When you are looking for a particular instrument's readings, you are looking for a device-and-metric pair.

Where you will meet this again

Device then metric is the selection path throughout the product. Adding a graph card, defining a calculation variable, and setting an alert threshold all begin by picking a device and then one of its metrics.

It is also the order to troubleshoot in. No data anywhere points at the connection. One instrument missing points at the device. A chart with an unexpected shape or scale usually points at the metric, either the wrong one selected or the wrong unit assumed.