The Google Kubernetes Engine docs have a section for Understanding cluster resource usage that explains how to have GKE send resource consumption data to BigQuery via the --resource-usage-bigquery-dataset flag.
However, there's also Stackdriver Kubernetes Engine Monitoring, which seems to accomplish the same thing, only by sending the data to Stackdriver.
Why are there two options? What are the differences between them?
Each product has a use case.
For example, if you would like to analyze data for a long period of time. You will likely choose GKE usage metering with BigQuery because the data has no retention period unlike Stackdriver monitoring which has a 6 week retention period.
In general, if you plan to export your Metric Data anywhere. You will need to pass through BigQuery.