Want this feature implemented, too? Please upvote by clicking Thumbs up!
(Posts as a reply won't be published in this feature request thread. Read Me!)
As a PRTG User, I want to see sensor/channel historical metrics aggregated by time slot, so I can have a better understanding of whether current sensor values are actually within normal/expected behaviour or not.
Details of User Story
Sensor historical data and charts give us point-in-time values only, with no direct way (through PRTG) to compare whether that value is within a normal/acceptable range for that point-in-time or not.
Our current workaround to this is to manually generate aggregates as needed:
- Generate historical data via PRTG for a sensor, export to CSV
- Using Excel, allocate each historical data point to a time slot we're interested in (for example, there are 96 15-minute slots in a 24-hour period)
- For all the channel values of a given time slot, we can now generate:
- Minimum value
- Average (mean) value
- Median value (requires Excel Power Pivot add-in, if using Pivot Tables)
- Maximum value
- Standard Deviation
The above process is quite time-consuming, and although can be semi-automated, requires us to use tools outside of PRTG to do a simple analysis of cases such as:
- CPU usage of 80% at time <x> - is this normal or not? Comparing to trends generated by above analysis, we can immediately tell if this is something to be concerned about
- Reports can be generated to see how our systems are behaving on average at different points in time of the day, days of the week, weeks of the month, etc
- Get a much clearer picture on whether something is going on, and also perform better capacity planning (especially from sensors looking at CPU, memory, bandwidth usage, disk I/O)
The above aggregates can be provided as additional columns to the historical sensor charts and/or can be a separate sensor reporting option which only generates aggregate data.
Nice to Have
Option to create a "rolling window" for aggregate generation; so for each point-in-time its corresponding aggregate data would be created using <x> data points in the same time slot prior to that data point
- Sensor historical charting should provide the aggregated values for each time interval alongside the point-in-time values which are currently provided
- Option to select how far back to look when generating aggregates (e.g. aggregate over 30 days, vs looking at last 24 hours of point-in-time data)