1833778 Members
2278 Online
110063 Solutions
New Discussion

vCPU and Entitlement

 
SOLVED
Go to solution
kemo
Trusted Contributor

vCPU and Entitlement

Hello

I've read a lot about vCPU and Entitlement. still confused all the way.

Can you please explain according to the below example.

I have 2 Physical CPU , each is dual core.

suppose I'll create 1 virtual machine with 1 vCPU and Entitlement 40% .

Q1: how many virtual machines i can create with this entitlement.

Q2: This 40% is a percentage of each CPU or each Core?
Q3: what is the maximum Entitlement which i can assign to any virtual machine? Can i assign for example 150% as Entitlement for one virtual machine?
Q4: Numbers of vCPUs must be less than or equal to number of physical CPUs or physical Cores ?

appreciate if you can answer the above queries.

Thanks
3 REPLIES 3
Eric SAUBIGNAC
Honored Contributor
Solution

Re: vCPU and Entitlement

Bonjour,


Q1 : Your question should be "how many virtual machines i can RUN..." because you can CREATE as many guest as you want.

Well : on your 4 core physical host, you can theorically run simultaneously 10 virtual machines with 1 vCPU / 40 % entitlement each. Total entitlement = 400 % wich doesn't exceed the whole processing power of the host

In fact its is not so simple, because you will start no more than 8 virtual machines in the configuration you have described.

Entitlement is a GUARENTEED % of processing power of a core. GUARANTEED means that a guest will not be started if the entitlement can't be honored by the host. If 8 VMs, with 1 vCPU and a 40% entitlement each, are started on your 4 cores host, you will get :

- core1 : vm1 and vm2 with 40% entitlement each --> 80%. So the host can only guarantees 20% for running an other VM on this core
- core2 : vm3 and vm4 with 40% entitlement each --> 80%. So the host can only guarantees 20% for running an other VM on this core
- core3 : vm5 and vm6 with 40% entitlement each --> 80%. So the host can only guarantees 20% for running an other VM on this core
- core4 : vm7 and vm8 with 40% entitlement each --> 80%. So the host can only guarantees 20% for running an other VM on this core

Q2 : ... of each core mapped to each vCPU of the guest

Q3 : 100% is the maximum entitlement. On a 1 vCPU guest it means you can have 100% of one physical core, on a 2 vCPU guest it means you can have 100% of 2 physical cores, and so on ...

Q4 : a vCPU relies on a physical core. Simply forget CPU on the host, and just think in term of core. So number of vCPU in a given guest must be less or equal than physical cores.


As a summary and a guideline, you must think to vCPU in term of a thread running at a time on a given physical core. Several vCPU of a given virtual machine can't run on the same physical core.

HTH

Eric
Emil Velez
Honored Contributor

Re: vCPU and Entitlement

2 cpus 2 cores

when you run top you see 4 cpus.

YOU CANNOT turn on hyperthreading with Integrity VM.

The entitlement is maximum entitlement so you can have more allocation than you have and it depends on what VMs are running.

hpvmcreate -e 20 -c 2
or
you can create VMs with -e -E to specify the number of physical CPUs allocated to VMs.

creates a 2 processor VM with 20 entitlement.

Your virtual machines can have as few as 1 CPU and as many as the amount of virtual CPUs as physical CPUs.

THe real limitation is memory since memory for running VMs is taken off the top.

integrity VM is very simple.

1. create the switch on lan0
2. boot the switch
3. create the VM
4. add a lan card to the vm on the switch
5. add a disk to the vm


hpvmnet -c -S switch1 -n 0
hpvmnet -b -S switch1
hpvmcreate -P vm1 -e 20
hpvmmodify -P vm1 -a network:lan:1,0,aabbccdd0005:vswitch:switch1
ioscan -funCdisk
hpvmmodify -P vm1 -a disk:scsi:0,0,4:disk:/dev/rdsk/c0t1d0

Now update /etc/bootptab with the mac address of the VM and boot it (assuming system has the ignite software)

hpvmconsole -P vm1
boot from the lan card.

kemo
Trusted Contributor

Re: vCPU and Entitlement

Thanks Emil Velez

But sorry your answer is far away from my query. but thanks anyway.