qbee allows to collect metrics from all remote edge devices. Both CPU load, memory, disk usage, bandwidth and many other parameters can be checked. This allows a detailed view into the system and allows to pro-actively see problems before they occur. All metrics can be constantly monitored and alerts can be initiated if thresholds are crossed. It is possible to browse back in time and adjust the time window of the graphs through the time selector. Metrics are available by clicking on a group in the device tree or alternatively from the "Devices" menu under the "Graphs" tab for a single device view.
By opening a rectangle with your mouse you can zoom in into interesting areas.
Double clicking into the graph allows you to get out of zoom mode.
In group mode multiple graphs are displayed in one graph window. Below the graph is a legend. Clicking on devices (or rx/tx) there toggles if the specific graph will be shown or not. Above all the graphs there is a time selector allowing to select the time period of interest.
Comparing different device metrics
By clicking on a group the devices with the highest values are computed and visualized. This is very helpful to find deviations in a group or compare different devices. Through this aggregation it is possible to see all devices with highest values in each category. This is very helpful to find deviations or debug problems.
This metric data collection helps to identify anomalies and deviations. We have seen both industrial applications or standard Raspberry Pi applications from a standard Raspbian install leak memory. Customers have also noticed that extensive debug log file logging was not disabled on devices that were deployed remotely. This could have caused a full crash. But with the remote metric collection it is possible to identify performance issues early on and secure and harden the IoT edge infrastructure.
Digging deeper into single device view: Below all graphs of a single device are shown. qbee transmits one value per agent run. So if the agent has a 60 minute run interval data points will spaced with 60 minutes. Moving above 24 hours even 5 minute intervals will be aggregated into larger buckets. This can change the way the bandwidth graph looks.
The last extension of our metric graphs is the bandwidth part. Here all interfaces are plotted both for rx and tx as well as a combination of the complete traffic. This helps to analyze and monitor system network behaviour in order to optimize or see deviations. Customers use this to estimate which data package they should purchase for mobile network connections and to identify applications that have unexpected bandwidth demands.