Trend CPU utilisation for a service

Question

Hey everyone,

Hoping you can assist. I would lke to show, graphically, the CPU utilsation for a VM or VMs contained within a service over a 6 month period. I would also like to overlay the average utilisation to demonstrate whether the utilisation has increased or decreased over this period.

Easy enough to do the first part, but do you know how I could also show the average utilisation over this period ? The only "average" overlay I can think of will show a flat average over the entire period. I would really like to demonstrate whether the average has been increasing or decreasing.

Cheers

DELL-Brian W · Accepted Answer

This is a good use case for a Derived Metric.

A Derived Metric is calculated based on one or more other metrics.

In this case, you could define a Time- or Schedule-driven Derived Metric that's scoped to a Service.

The calculation to get the average CPU utilization for all hosts in the Service would look something like:

avg(#utilization from $scope.hosts.cpus for 30 minutes#)

Once you've created this metric, you will be able to chart this value over time and monitor it for changes.

There's an example of how to create a Derived Metric in How to create a Derived Metric (to count powered on VMs).

Regards,

Brian Wheeldon

mrs1 · Answer

Hey Brian, Thanks for that, it works like a treat !  Has anyone ever asked whether you could use derived metrics to provide a capacity figure based on the number of VMs remaining ?

DELL-Brian W · Answer

I'm not sure what you mean by 'a capacity figure based on the number of VMs remaining'. Can you clarify? Which metric or metrics would this be derived from? Regards, Brian Wheeldon

mrs1 · Answer

Brian,

Sorry I meant the reverse of that...

Could you create a derived materic which represents the remaining environemtn capacity in terms of numer of VMs that could be deployed?

Eg unsued memory & unused cpu / VM spec 4Gb Ram 1 vCPU = VMs available value.

DELL-Brian W · Answer

This sounds like something that's already handled by the Capacity module.

The screenshot posted in the thread VMworld 2011: vFoglight User created carts/modules? clearly shows the VMs available.on the VMWare Environment/Capacity Management tabl.

Nevertheless, the derived metric you've described would look something like:

memory_remaining = unused_memory / memory_required_for_one_vm

cpu_remaining = unused_cpu / cpu_required_for_one_vm

vms_available = min(memory_remaining, cpu_remaining)

So if you can decide what the four input variables are, then you can calculate this derived metric.

Regards,

Brian Wheeldon

sundling · Answer

Capacity Management is available out of the box in vFoglight 6.6 and above. It tells you Available VMs based on CPU, Memory and Disk. See below screenshot:

Virtualization Infrastructure Management

Trend CPU utilisation for a service

Was this post helpful?