Hi, we will deploy a number of vSphere servers with the following spec:
These servers will host a number of VMs with an identical build:
Can you detail the maximum number of VMs that I can run on the physical server? and also how you calculate the figure.
Also, whether there is difference if Hyperthreading is enabled.
Thanks
this all depends on workloads. Without testing no one will be able to give you and exact number. How many IOPS do you expect? Recomendations are 4 - 8 VM's per physical core
if there is already workload running, you can use vmware capacityIQ to predict how much more workloads you can put in ..
I don't know the IOPs at present as nothing is built and this is a planning stage. Also, I'm aware VMware suggests best practice of 6 VMs per core.
At this stage need to gauge a rough idea of VMs per host.
nothing is built at present and the detail I've provided is the requirement from which need to assess the VM capacity
Also, I'm aware VMware suggests best practice of 6 VMs per core.
Recommendation is from 4 vCPU to 6vCPU per physical core,
At this stage need to gauge a rough idea of VMs per host.
I that case you have to play with assumptions:
assumptions:
now do the math
Physical host specs
2 quad core (8 cores) and 96GB RAM
(4vCPU * 8 pCPU) * 1vCPU = 32 VM's - CPU oversubscription 400%
3GB RAM * 23 VMs = 99GB - oversubscription 101%
As you can see, key is average VM size.
Good luck
Without data, theres no efficient way to tell besides a guess.
Some people get 1 VM/host (huge DB servers), some get 100 (in VDI scenarios) on the same server. Its 100% workload dependent.
I've supplied the detail as:
vSphere Host:
These servers will host a number of VMs with an identical build:
Assuming 3vCPU per Core, the vCPU per host is 3vCPU*12 pCPU =36vCPU
Therefore, should be feasible for at least 18VMs and memory requirement is 2*18=36GB. Of course, need to allow for overall loading of vSphere server and not utilise all capacity - say max utilisation is 75%, then in theory should be OK for at least 12-13 VMs - do you agree?
joeflint wrote:
I've supplied the detail as:
vSphere Host:
- 72GB Memory
- 2 x Intel Xeon X5680 6C/12T 3.33 GHz - the CPU's are 6 Core i.e. 2x6 = 12 cores
These servers will host a number of VMs with an identical build:
- 2GB memory
- 2 vCPU (requirement is 2 GHz per vCPU)
Assuming 3vCPU per Core, the vCPU per host is 3vCPU*12 pCPU =36vCPU
Therefore, should be feasible for at least 18VMs and memory requirement is 2*18=36GB. Of course, need to allow for overall loading of vSphere server and not utilise all capacity - say max utilisation is 75%, then in theory should be OK for at least 12-13 VMs - do you agree?
Well, your calculations are very conservative, believe me you can squize more from that hardware
You should enable HT on your machine then you will have 24 cores
If you have requirments 2GHz per VM then your VM specs should looks like
1vCPU, 2GB RAM
Following your HW specification I would do in that way
(3vCPU * 12 pCPU)*1vCPU = 36 vCPU
36vCPU * 2GB RAM = 72GB
Following your inforamtion that you will have all VM's in this same specification ( OS as well) you will get a lot of from VMware memory saving techniques
If you wanna be super safe you can say 25 - 30 VM's per host