Underperforming Network Benchmarks

Issue

• After running some iperf3 bandwidth tests between OpenShift nodes and the results are below expectations
• iperf3 was used between 2 OpenShift Container Platform workers on the same Red Hat Virtualization node
• The network uses an MTU of 9000 and the MTU of the SDN is 8950.
• Questions
  • Are there default constraints put in place to limit networking performance in Pods?
  • Any reason why there's a disparity when comparing the host network performance and the SDN?
  • Any tweaks to achieve higher speeds within the SDN and container?