When it comes to defining logical compute design we need to consider following aspects,
1. Total requirements in terms of CPU and Memory
2. Thresholds for maximum utilization
3. If any growth to be accounted for?
For example; We have a functional requirement of 200000 MHz of cpu and 300000 MB of RAM. There should be a consideration for 20% growth in the workload. The infra should provide N+1 redundancy.
Host selection: This will depend on the feature requirement, preference of the vendor or specific make / model. For simplicity of calculations, i am not considering any specific models or vendors.
No of CPU (sockets) per host 2
Cores per CPU 8
MHz per CPU Core 3000
MHz per CPU 24000
MHz per Host 48000
Maximum Utilization Threshold 85% = 40800 MHz
Total RAM per host in MB 98304
Maximum Utilization Threshold 85% = 83558 MB
To Satisfy CPU requirements,
Functional Requirement = 200000 MHz
Available CPU per host = 40800 MHz
Number of hosts required to satisfy this requirement : 200000/40800 = 5 hosts (actual figure 4.90)
To Satisfy Memory requirements,
Functional Requirement = 300000 MB
Available RAM per host = 83558 MB
Number of hosts required to satisfy this requirement : 300000/83558 = 4 Hosts (actual figure 3.59)
Now we need to keep in mind that although 3 hosts can satisfy RAM requirement, they can not satisfy the CPU requirements. Hence we need to go with the figure that will satisfy both requirements i.e. 5 hosts.
Now we need to calculate for Growth and redundancy.
Number of ESXi hosts required = 5
Growth 20% CPU = 40000 MHz
Growth 20% RAM = 60000 MB
Redundancy = N+1
Final host requirement : 5 hosts (Required to satisfy the requirement) + 1 host (20% growth) + 1 host (N+1) = 7 Hosts