I want to measure the Windows Container performance and met a strange problem.
I have two servers installing Windows Server 2016(Standard Edition) and use NTttcp measuring the bindwidth of two servers. If the server don’t install the Docker(17.03.01-ee-3), the bandwidth of two servers is 17-18Gbps. However, the bandwidth of two servers is 6Gbps when the server has installed Docker. I have no idea why the Docker will affect the bandwidth of the bare metal.
Moreover, if I make the server(installed docker) as sender(the receiver doesn’t install docker), the bandwidth will decrease. But if i make the server(installed docker) as receiver(the sender doesn’t install docker), the bandwidth isn’t affected. I am so confused…
Does anybody meet the same question? Why does the bare metal’s bandwidth decrease after installing Docker on Windows Server 2016?