Why does vNIC utilization differs from expected values?

[ucsadminbi]

Hello community,

I came up with an observation I could use some help to understand it.

Scenario:

On our blades runs VMware ESXi with some Linux VMs in it. Using iperf3 to test the bandwith between to VMs running on seperate blades in the same chassis. Iperf3 shows a bandwith of 8 Gbit/s. The vSphere Client shows a bandwith of 8 Gbit/s as well for the used vmnic. But looking at the vNIC of the blade in UCSM it shows only 4,4 Gbit/s.

I don't understand the reason for that? Does someone of you has an explanation for that?

Regards, 

Joerg

[Kirk J]

I wonder if the sampling time block is different on FIs (i.e. 15 min, 30 min etc) vs an instant or 5 min average you might see on VMnic on esxi?

Obviously, if you left this running for an hour, I would expect them all to have roughly the same #s.

Kirk...

[ucsadminbi]

Hi Kirk,

Thanks for your reply.

Yesterday, I had the chance to run the same test for an hour. But the bandwith won't add up. It still around 6.7-6.9 GBit/s in vSphere Client and guest OS but only around 3.9-4.4 in vNIC in UCSM. And I have no clue about the How or Why, yet. I guess it could have to do with different timeframs between different datapoints here, but it seems odd that there is such a huge gap.

Regards, Joerg