I am in process of implementing a new system where we need to store time series performance metrics. The metrics collected provide us information on how a specific robotic device is performing. Metrics include distance traveled, amount of liquid transferred, amount of time waiting, amount of time moving.... All of this is time series data, but I would say very little of it is "predictable". Meaning a robot can take 4 minutes to move one time, next time it might be 90 minute, next time it might be 2 hours. While there are upper / lower bounds (can't take 0 seconds to move and probably won't take days) the window seems to be significant. Much of the data falls into this scenario, so given this I'm wondering how efficient it's actually going to be storing it in a historian. If my compression/exception windows have to be so wide, will I just end up storing all the data and therefor defeating the purpose? One of our primary objects is to do robot by robot comparisons, robot group metrics ..... Little focus on individual robot performance. What's the best way to do multi robot comparison in PI?