From a monitoring perspective getting the overall cpu and memory should be sufficent for alarming. The extensive view becomes the troubleshooting process during an issue.
Also the most critical processes will generate alarms and traps when errors occur so these would be consumed by the monitoring system to flag the problems rather than the processes directly.