I have read from Vmware that a disadvantage of the IP hash load balancing option is increased CPU overhead in the host.
That would come from having to do a hash of some portion of the destination and source IP adress of each frame to know which link to put the frame on. That will not be necessary on the Port or MAC based since it is "hard coded" per VM.
My question is, how big is this CPU overhead? On a fairly modern CPU, would these calculations be noticable, as in less CPU time available for the VM guests?
I am not looking for a exact answer as in "it takes 2,9 % cpu time", just what your experiences are?
If you won't run your ESX with 80+% CPU load then overhaed is negligible.
---
MCSA, MCTS Hyper-V, VCP 3/4, VMware vExpert
If you won't run your ESX with 80+% CPU load then overhaed is negligible.
---
MCSA, MCTS Hyper-V, VCP 3/4, VMware vExpert
Thanks, so it should not be a problem likely? Does it matter if having just 2 Gb VMNICs teamed or having, say, 10 Gb nics in some different teams for VMs, Ip storage and similar? That is, if having multiple teams - would the CPU overhead be higher?
You can always change settings and measure the difference.
---
MCSA, MCTS Hyper-V, VCP 3/4, VMware vExpert
I have not really got that enviroment, was only curious if there was any known problems with that kind of setup.