The 'iperf' tool is used to test network performance. The 'iperf' test perform multiple iterations to bandwidth between two virtual machines.
The results of the 'iperf' tool may show high number of retries - indicating network performance issue.
vSphere ESX
These retries indicate failure of packet receiving on destination side. A number of factors may cause the higher number of retries.
Additional data should be collected included networking stats from the switchport associated with the virtual machine.
Review networking stats for the vNIC attached switchport using following command
vsish -e get /net/portsets/<SwitchName>/ports/<PortNum>/stats
In case if the output of above command shows high number of Rx packets drop then
#1. Review network adapter type attached to the virtual machine
'vmxnet3' adapter provides best performance and is installed with the installation of vmware tools.
#2. Review 'rxSummary' stats, to determine the status of 'ring buffers'
vsish -e get /net/prtsets/<SwitchName>/ports/<PortNum>/vmxnet3/rxSummary
In case the output stats contains 'running out of buffer' more than zero (0) then it indicates the exhaustion of received buffers. The increase the 'ring buffers' for the guest OS will help to avoid the exhaustion of the buffers.
For additional information related to performance tunning
Performance Tuning for Latency-Sensitive Workloads VMware vSphere 8