VMware capacity planning, with failure simulation.

Is the cluster enough if a host fails? Do ten new VMs still fit? Which VMs are too large, and when will storage be full? Spherifyer answers this from the real data of your vCenter and the history of the last weeks, not from a hand-made spreadsheet.

See the failure simulation Documentation

01HA reserveCan the cluster absorb a failure?

N+1 means: the largest host may fail.

vSphere HA restarts the VMs of a failed host on the remaining ones, but only if there is enough free RAM and CPU. Spherifyer calculates this for every cluster: what is left when the largest host is missing? If it is not enough or gets tight, a warning with a recommendation appears in the dashboard.

Forecast: Usage per cluster: CPU, RAM, storage and vCPU per core, plus when storage will be full at the growth of the last 30 days.
RAM: Warning when the RAM of the remaining hosts is not enough or gets tight after the largest one fails.
CPU: Warning when the CPU of the remaining hosts cannot carry the load.
Overcommit: vCPUs per logical and per physical core, with a threshold, plus CPU ready per vCPU (warning from 5 %, critical from 10 %).

02What-ifPlay through failures before they happen

Fail hosts, datastores or whole sites with a click.

The simulation shows the hosts with their VMs. When a host fails, its VMs visibly move to the remaining ones, just as HA would restart them. Whatever finds no room is listed under “Cannot start”. VMs on a failed datastore or on a local datastore of the host stay down. When a whole site fails, its VMs start at the other site as far as RAM is free there.

Examples: Largest host, two hosts, half the cluster, host-by-host maintenance, datastore failure, or hosts of your choice.
New VMs: Add planned VMs and see how many VMs of that size still fit afterwards.
Verdict: A result with reasons and limits, plus a table of the affected clusters.

03ReclamationFree up capacity before buying

Often the capacity is already there, just poorly distributed.

Before ordering a new host, it pays to look at what is allocated but unused: VMs that have been off for weeks, idle VMs, old snapshots and VMs with more vCPUs or RAM than they need. Spherifyer assesses this from the history, not from a single measurement.

Reclaim: VMs powered off for 30 days, VMs idle for 14 days and snapshots older than 30 days, with the total storage, vCPUs and RAM. With monthly prices also the estimated saving.
Rightsizing: High and low CPU and RAM usage per VM, as average and P95 over the measurements of several days.
License requirements: Processors and cores per datacenter, as the basis for planning your VMware licenses.

04FAQFrequently asked questions about capacity planning

What does N+1 mean for a VMware cluster?

The cluster has one more host than the load needs. If a host fails, even the largest one, its VMs restart on the remaining ones. Spherifyer checks this for every cluster and warns when RAM or CPU will not be enough afterwards.

How does Spherifyer simulate a host failure?

With the real VMs and hosts from your vCenter. The VMs of the failed host are distributed to the remaining hosts of their cluster as HA would, as far as RAM is free. Whatever finds no room is “cannot start”. The simulation changes nothing in vCenter.

Can I play through a site failure?

Yes. In the topology you let a whole datacenter fail. Its VMs then start in the free RAM of the other datacenters, as far as it lasts. A verdict with reasons is shown above.

How do I find oversized VMs?

Spherifyer evaluates the CPU and RAM usage of every VM over several days, as average and P95. If a VM stays low permanently, it suggests fewer vCPUs or less RAM; for the CPU it also takes the CPU ready time into account.

Does Spherifyer help with VMware license planning?

Spherifyer shows how many processors and cores each datacenter contains. That is the basis for licensing by cores. The license terms themselves are set by the vendor; Spherifyer does not calculate prices.