Reclaimable Capacity

[StuWe]

Hi All.

I have been messing around with the vRealize Operations Manager to see if there is value of this in our organization. Of course management always wants a break down of the benefits that it will add and how this could potentially save money, capex and opex. I am currently looking at the Reclaimable Capacity report and Analyses. I can see we have a lot of cores and Memory that we could reclaim, but I am struggling to equate that to a $ Value.

From what I understand, if I were to reclaim this capacity, I would only be utilising the cluster more effectively, I should see drops in Balooning and CPU Wait time for those resources that are needlessly requesting cores and reclaiming memory from VM's. Our overall usage of our the Cluster would not change much? I could not work out for instance that I could drop a host out of the cluster from the reclaimed space.

How have others worked this out to justify purchase?

Regards,

Stuart

[daphnissov]

Something that would help you is to integrate vRealize Business for Cloud into the mix. You can now connect the two together. vRBC will show you costs for these wasteful machines based on your underlying infrastructure. Also, use the capacity planning ability in vROps to model what it would look like if you *removed* a host, then paired with vRBC you can really start to understand how rightsizing your VMs leads to removing excess host capacity which leads to $$ savings.

[sxnxr]

This is a big question that is hard for us to answer as we dont know your capacity policies.

For us we use an allocation based model that allows us to say we have x deployments of 4 x 32 VMs in a cluster. We then set an over commit ratio that is close to maxing out the cluster (we dedicate a host for HA failover in vcenter that way vrops works out the HA buffers no matter how many hosts are in the cluster) . What we class as maxing out a cluster may be different to you. In our case we look at ready and contention as well. By having the ratio i know my cluster has the ability to host x number of vCPUs.

I then look at the reclaimable for that cluster and then work out now many free vCPUs i would have if we reclaimed them and that would save me buying x number of hosts. My formula would look something like: reclaimable vcpu count / (New hosts cores x over commit ratio) = how many hosts i would save * the cost of a new host

Depending on the size of your environment this may be 1 host or hundreds. One thing i would say is dont trust the memory stat in reclaimable. It is based of active memory on the host and does not take into consideration application cache.

[StuWe]

Thanks daphnissov,

I am busy playuing around with the projects to see if I can isolate prooblematic/demanding VM, because if I take the current value I says I require a new host. If those issues are resolved and the project no longer needs a new host, I have effectively saved Money

[StuWe]

Thanks sxnxr,

I see how you have structured your own environment. I think we are still defining our own policy, which does make this complicated. Your input however does give me a new train of thought to following during this process. Much appreciated.