How is calculated the "Memory Overall Workload" metric?

[Nico06]

Hello,

My question may seem stupid, but I ask it because I'm in front of the following case:

I have an host in my vCenter that I monitor with vCops and the vSphere Client, and a few hours ago, I have seen an alert from vSphere telling that my host memory usage was high. And when I watched, in fact, 9.5GB of 10GB where used.

BUT... A few second after, I have seen on my vCops web interface that the "Memory Overall Workload (%)" was about 36%, whereas the "Memory usage" (still on vCops) was about 9,6 GB.

Consequently, I think that "Memory Overall Workload" is not equal to "Memory Usage" / "Total Memory Capacity" * 100

So, how is it calculated? What this metric really represents?

Thank you in advance,

Nicolas

[emma234]

Hi Nico06

Welcome to the community .

Hope below pdf will help you and give more idea.

http://www.vmware.com/pdf/vcops-standard10-install-admin-guide.pdf.

"He Conquers, Who Conquers Himself".

[Nico06]

Hello, thank you for your answer.

Unfortunately, I have already read this document, and the other for the version 5.

And unless I'm mistaken, I didn't find information which can help me

I think the "Memory Overall Workload" metric is the result of "Memory Machine Demand" / "Total Memory Capacity" * 100 (because the 2 curves are similars), but I don't really understand the meaning of the "Memory Machine Demand" metric.

If I get an alert with vSphere because the occupied memory of my host is 95%, logically, I should have the same on vCops, no?

Regards,

Nicolas

[beckham007fifa]

Nico06 wrote:

Hello, thank you for your answer.

Unfortunately, I have already read this document, and the other for the version 5.

And unless I'm mistaken, I didn't find information which can help me

I think the "Memory Overall Workload" metric is the result of "Memory Machine Demand" / "Total Memory Capacity" * 100 (because the 2 curves are similars), but I don't really understand the meaning of the "Memory Machine Demand" metric.

If I get an alert with vSphere because the occupied memory of my host is 95%, logically, I should have the same on vCops, no?

Regards,

Nicolas

Workload, simply put, is a measurement showing the ratio of the resource demand of a virtual object (VM, ESX, Cluster, etc) versus the amount of resources it can obtain.

If the Workload score is low (e.g. 20) an object has plenty of resources at its disposal.  The higher the Workload score the closer a virtual object gets to running out of its necessary resources.  A Workload score above 100 (yes, it can be >100) means the virtual object is starving for resources (CPU, Memory, Network I/O or Storage I/O).

[beckham007fifa]

Nico06 wrote:

Hello,

A few second after, I have seen on my vCops web interface that the "Memory Overall Workload (%)" was about 36%, whereas the "Memory usage" (still on vCops) was about 9,6 GB.

Consequently, I think that "Memory Overall Workload" is not equal to "Memory Usage" / "Total Memory Capacity" * 100

So, how is it calculated? What this metric really represents?

Thank you in advance,

Nicolas

how can you say that the host memory usage was 9.6Gb? could you elaborate this.

[Nico06]

By reading the "Memory Usage (KB)" metric in the Operations > Details pane of my host in vCops, and the "summary" pane in the vSphere Client.

Ok, so "Overall Workload" for a resource is not its usage, but its disponibility in case of demand?

[beckham007fifa]

Nico06 wrote:

By reading the "Memory Usage (KB)" metric in the Operations > Details pane of my host in vCops, and the "summary" pane in the vSphere Client.

Ok, so "Overall Workload" for a resource is not its usage, but its disponibility in case of demand? - YES,

workload = demand of the virtual object/amount of resources it can obtain,

lesser value means resource available is more than the demand.

But your figures which you have provide in the first comment is pretty subtle to explain and support. The values are not getting substituted in the equation that might be the case when you had a glance on your setup. I would say you should check again now and the values should be in proportion else we are missing something.

[Iwan_Rahabok]

Workload is not related to Usage. Usage is mapped to Consumed. Usage is mapped to Active at VM level, but to Consumed as ESXi level.

Workload is based Machine Demand. I'm not 100% sure what the formula for Machine Demand is. It contains Active as the pattern is similar, but definitely not just Active as the value is _always_ higher. I'm using vC Ops 5.8.1 and I checked on a few ESXi 5.5.

Hope that helps

e1