As you monitor a service employing n2-standard-2 Compute Engine instances for delivering large files, users complain of slow download speeds. Cloud Monitoring indicates that VMs are operating at maximum network throughput. How can you improve network throughput performance?