regarding mentioned "view for last point" - more precisely querying metric values for given set of dimensions that have most recent timestamp: how to achieve it? can it be created in PQL or schema or maybe requires some C++ plugin?

What’s the unique id in the dimension - device id or tag_id

simplified example: there is more than one dimension for example: device_group, device_type, device_id and possibly more than one metric, for example cpu1_usage, cpu2_usage and obviously time. 1. for each set of dimensions (device_group, device_type, device_id) I would need to know most recent cpu1_usage, cpu2_usage (and time when it was measured). 2. calculate the average cpu1_usage from those latest values for given device_group 3. calculate the average cpu1_usage from those latest values for given device_type in above calculations older values of metrics for same dimensions (device_group, device_type, device_id) are not relevant - similar to in-place update in RDBMS for the key (device_group, device_type, device_id)

Got it, you need the latest value to be maintained for a given set of dimensions.