How do you calculate the potential metrics generated?

[Hurleyinnh]

I'm trying to determine what the requirements will be for me to deploy vCenter Operations Manager Enterprise 5.

How do you calculate the metrics generated? I have 6 vCenters in 4 locations, 2 in US, 2 international, with about 160 hosts running 2,400 vm's.

Is there a formula for this?

I don't want to build too small and need to be able to expand for about a total of 200 hosts or more in the future running maybe 3,000 vm's or more.

Thank you for some insight into this ... Jim

[admin · Accepted solution]

You can use the following as an estimate:  1100 metrics/ host and 400 metrics / vm      So, for 200 hosts and 3000 vms you will have (approximately) 220000 + 1200000 = 1.4 M metrics.

Hope that helps.

-dw

[wharlie]

From "deployment and confiuration guide"

http://www.vmware.com/pdf/vcops-5-installation-guide.pdf

Each virtual machine has 250 to 300 metrics.
Each ESX host has 1,200 to 1,500 metrics.

(#VMs x 300 metrics) + (#Hosts x 1,500 metrics) = total metrics

In your case:

2400 x 300 + 160 x 1500 = 960,000

Also for storage:

Assuming

Metric collection frequency (every 5 minutes)

Metric retention period (default of 6 months)

Metric storage requirement (16 bytes per metric)

Using these assumptions it is possible to determine the amount of storage capacity required.  This is calculated as follows:

(#Metrics x 12 collections per hour x 4320 hours x 16 bytes)/(1024^3) = Storage (GB)

It is possible to determine the amount of storage I/O required.  This is calculated as follows:

(#Metrics/300 sec per collection period) = Storage I/O (iops)

These metrics are related specifically to the use of vCops in conjunction with the vCenter Server adapter and do not include the use of additional external adapters.  If there is a requirement to collect further data from additional adapters then there is a requirement to understand the total number of metrics this is likely to introduce.  For example if a SCOM adapter were also introduced it may collect an additional 250 metrics per guest operating system that would need to be factored into the above equation.