How would you monitor resource utilization (CPU, memory, disk I/O) of a running data pipeline, and what metrics indicate a bottleneck?
