vCheck 5 "Capacity Planner Info"

[livinma]

Hey All -

I recently starting diving into the world of powercli and reading a lot of posts from Luc and Al. I came across the vCheck script a few months ago and briefly looked at it. I am now tasked with coming up with some capacity information for my senior leadership and it seems that the vCheck script has some valuable information in it. My problem is under the "Capacity Planner Info" I am not quite sure what it is telling me. I really would love to see if this information will help out. I have attached a snapshot below of my results but unclear how to act on it. A bit of clarification would help so I can finish my report this weekend.

Regards,

Mark

[LucD · Accepted solution]

In the vCheck report Alan does a theoretical calculation of the guest capacity, cpu- and mem-based,  for a cluster.

The formula is not too difficult to understand:

• the ' real' cpu and mem capacity of the cluster. This is the number of hosts in the cluster minus the failover level you specified in the cluster settings. For example (I only show cpu, mem is similar) :
  • A cluster has 8 nodes.
  • Each node has 4 x 3 Ghz = 12 GHZ cpu capacity. The cluster has a total cpu capacity of 96 GHz.
  • The cluster failover level is defined as 2 nodes
  • The script calculates the 'real' cpu capacity as: 96 GHz - ((96 GHz * 2) / 😎 = 72 GHz
• the average cpu usage of the cluster over the last 7 days.
  • This uses the Get-Stat cmdlet with the cpu.usage.averagemhz metric.
• the number of VMs on the cluster
• the average cpu usage of the top cpu-using VMs
  • It calculates the average cpu usage of all VMs on the cluster
  • It selects the VMs that use more than this average cpu usage
  • It calculates the average cpu usage of these top cpu-using VMs
• a theoretical cpu usage value for a single VM, by dividing the above 'average cpu usage of the top cpu-using VMs' by the number of VMs on the cluster
• Finally, it takes the cluster's 'real' cpu capacity, substracts the 'average cpu usage of the top cpu-using VMs' and divides this by the 'theoretical cpu usage value for a single VM'

Note that this is an indication, and not an exact science, of what capacity you have left in your cluster.

There are other formulas in use to calculate such indicators.

[livinma]

This is great LucD. I truly appreciate the quick response. I will now go back and utilize the information you provided to see what I can come up with. Question ? You mentioned other formulas. Do you have a link of these formulas that I can utilize to calculate my capacity.

Regards

Mark

[LucD]

I'm afraid that these other formulas are more complex than the one from Alan's vCheck script.

A basic vSphere cluster concept (when using HA), that is not reflected in this vCheck calculation is the 'slot' concept.

For more information I point to Duncan's excellent HA Deepdive posts.

And even better, get Duncan's and Frank's latest clustering book.

Alan has a script to calculate the slots, see his HA Slot Size Information post.

Don't forget that although CPU and memory are the most important, there are other factors, like storage and network, that can influence the potential capacity of your cluster.

The golden rule I use; do a calculation of the potential capacity of your clusters as a guideline, but measure, measure and measure to see what is actually happening.

With the Get-Stat cmdlet you can collect a wealth of information, and by extrapolating you can make rather good predictions for the next months.

[livinma]

Luc -

Thanks so much for the additional information you have provided. This is certainly getting us 1 step closer to a process we can follow for our capacity management.

Regards