I had a client recently that had brought a small company in to P2V a small set of infrastructure. I was brought in to “fix” the issues that they left.
They made a small mistake that made me have to do some reading.
They were using Cisco EtherChannel to aggregate 4 GBE ports. This is fine and is an accepted way of increasing your bandwidth however they were not seeing the performance benefits that it should have brought. After some investigation I noticed that only one of the NIC on the ESX server was actually being used with any great demand.
Reading the below article led me to the load balancing option in the vswitch.
They had it set to “Route based on the originating virtual port ID”. The load balancing needs to be “Route Based on IP HASH”.
Also do not use beacon probing with IP HASH load balancing and do not configure standby up-links with IP HASH load balancing.
Once this change was made it load balanced as you would have expected (once the ESX network services had been restarted).