Network IO

[JimKnopf99]

Hi,

i have question about Network IO rate measuring.

Currently, i have an vm with high network IO. But it seems to be, that the calculation of this value is grounded to be an 100 MBit Network Card.

Why is it not using an 1 GB or more? The VM is using an VMXNET3 adapter.

Frank

[TomOtto]

Hi Frank

Check out the below thread, think it will answer your question:

Network I/O 100% Utilization

Also remember that there is a difference between Mbit and KB.

Regards

Tom Otomanski

[TomOtto]

One last thing, if you click on the "What does this mean?" link you'll find the below for Estimated Capacity:

Estimated Capacity is an estimate of the I/O capacity of the physical NICs, network switches, and endpoints based on the observed behavior of network I/O.

Regards

Tom Otomanski

[JimKnopf99]

Hi Tom,

ok. but one question remains. The Health status also depends on this value. If the Network IO rate is high, the Health value is getting lower. Therefor i receive an alert that my VM is consuming to much Network IO.

But my underlying host has a 10 GB Network connection and there are not many other vm´s running.

I thought, if the calculated size is is bigger, the IO weight is smaller and the Health is better.

Or am i missing something

Frank

[mark_j]

vC Ops uses a concept of "maxObserved". For as long as the vCenter adapter is running, it keeps a high watermark for certain values, calling it "maxObserved". This derived metrics is also visible when you look at a VM's network adapter attribute group. This derived metric is treated as the 'real-world' ceiling on a VM NIC's capacity. Given the variability in every network infrastructure's capability, this real-world ceiling is the most efficient way to have a solution that always gives you a logical answer. Over time, that maxObserved value will become truer as the capabilities of the environment are observed and effective maximums identified. This meets the need of defining network workload as a percentage of a it's overall capability, and even does so without knowing the network equipment+configuration that is outside vC Op's purview of monitoring.