How can you improve network throughput performance for n2-standard-2 Compute Engine instances serving large files when Cloud Monitoring shows they are operating at peak network capacity and users report slow downloads?