How would you optimize a slow-running SQL query that aggregates millions of rows of hardware telemetry data for Peloton interactive?
