vSphere networking performance counters - Teddy bear use of flow compared to its use

I am responsible for a vSphere 5.1 environment that uses NFS data warehouses and I'm trying to collect metrics on how much bandwidth is used to access NFS data stores.  When I did it first a week ago, I ran esxtop and looked at network and sort by Mbps passed parameters and what he was able to indicate what vmnic NFS vmkernel was linked to.  I then took a look at the usage meter vmnic for a total of send/receive KB/s.  During my research, I ran across another counter: "use of flow rates for NFS Teddy."  Here's the problem, when we look at the use and Teddy flow using the counters from NFS on the same vmnic over the same period of time, the use of flow for meter NFS shows bandwidth much higher than that of the meter use (see image).  What gives?

What I forgot to mention, is that measures to counter throughput.usage.nfs report a higher flow rate than the physical NETWORK adapter can handle.  The highest Summit on the graph shows about a flow of 13 GB/s on a network card 1 GB.  I spoke with a VMware engineer friend and he said it looks like the meter is mislabeled in vCenter.  I verified this by looking at the use of the NFS in vCenter Operations Manager metric and compared to the metric of its use on the same physical NETWORK adapter and the metric NFS for the NIC I was looking at showed approximately 23.8 MB/s (if I assume that the unit of measure of labelling was supposed to be B/s and not KB/s) and usage was about 24.2 MB / s.  The metric of higher use is due to the fact that the environment is badly designed by my predecessor and the NFS Sharing vmkernel rising with VM networks (something that I am trying to solve).

Tags: VMware

Similar Questions

Maybe you are looking for