As a basis for an upcoming post on splitting vmkernel traffic across over layer 3 boundaries I wanted to describe how vmkernel traffic is routed on an ESX host. There seems to be a lot of confusion in this area and hopefully this will help to clear it up. If you need a refresher on IP addresses, network masks, or subnets check out this Cisco article. Directly Connected Networks If a host is directly connected to a subnet it will use that interface to talk to devices in that subnet.