In a production cluster with 3 executor nodes where the driver and executors use the same VM type, which Ganglia Metrics indicator would identify a performance bottleneck due to driver-side code execution?