When you hover over the information icons in the title you can see the following tooltips:

• For the left column: “The metrics represent the most recent 50/95 percentile execution times of calls from the connector to your historian. Plot and index calls are calculated over the last hour, sync calls over the last 24 hours.”
  
  So the left column shows how long it takes for the TrendMiner connector to receive a response from the data source when it sends out a request.
   
• For the right column: “The metrics represent the most recent 50/95 percentile of calls total execution times from TrendMiner to your historian (data include the execution times from the connector to the historian). Plot and index operations are calculated over the last hour, sync operations over the last 24 hours.”
  
  So the right column shows how long it takes for the TrendMiner appliance to receive a response from the data source when it sends out a request. Note that if the data source is connected via a TrendMiner connector the timings from the right column include the timings of the left column and therefor will always be bigger.

The timings are split up between:

• index: all data request calls which are done by TrendMiner to build up the index for a tag. These calls are done only once per tag per time period (unless a tag is reindexed manually).
• plot: all data request calls which are done by TrendMine to fetch live data. These responses are not saved in TrendMiner and the requests are typically for smaller time periods. Live data can be disabled for a specific data source in which case no data is available.
• sync: calls to update the list of available tags in TrendMiner. These calls only happen once every 24 hours by default and therefor these timings have minimal impact on the performance.

Q: what are reasonable values?

A: this depends on the requested time frame. The metrics do not distinct between small timeframes (e.g. 1 hour) or large timeframes (e.g. 1 month) but of course the timings might differ significantly. But overall we consider a good value to be under 1 second (for plot and index calls). Timings up to a few seconds are acceptable but timings above 10s could indicate performance issues. TrendMiner has some configuration options to mitigate the impact of a slow data source or to improve performance by tweaking the connection parameters but this depends on the type of data source. For deeper investigation reach out to support via support@trendminer.com

Q: Over what time are these captured? (are these reset at some time?)

A: these timings (plot and index calls) are covering a rolling window of 1 hour so they only represent the last hour. If an optimization is done it can take up to an hour before the full result is visible in the metrics screen.