Collecting a very large number of datapoints

Are referring to P-states on the CPU which controls its running frequency? Would that need to be a separate datapoint for each state? I would think it would just be one PState datapoint that would store 0, 1, 2, etc. For 30 servers with 80 cores that would 2400 instances. Or if you create a device per host, would only be 80 instances.

Otherwise it does sounds like you'll have to break it up into multiple devices unless you can accept some sort of summary collection. Even if you got this working, you might run into the limit of the metrics-per-device (see https://www.logicmonitor.com/support/logicmonitor-limits-quotas-and-constraints) limit which I haven't seen myself but this sounds like it would hit it.

You said you need to collect data so I assume just setting up the Datasource to only create instances during an alert event is not enough. Otherwise you can have the AD script only create the instance when it's important but that also means you only can do that at 15-minute intervals and not really useful for graphing data.

If you do need to collapse into multiple datapoints, you can just create as many as the most you need. So if there is 16 at most, create the 16 datapoints but if the CPU doesn't have that many you can just skip the ones that don't apply. They will report NaN and not even show in a graph.