Query on VM

[raiden]

Guys i got this question from another post which was left unanswered. Now even i have the same query . The post question was as follows

We have already implemented HP IVM in rx6600 server(VM-HOST OS: HP-UX V3 VSE).I have a query on vCPU. The query is, say we have "12" core physical CPU.I have assigned 8 vCPU into VM machine "server1" & "server2" with 40% entitlement. So the summation of Entitlement=(40+40)=80% & this 80% entitlement comes from 8 core Physical server. Now i want to create VM machine "server3" & server4" with 5 vCPU
& 30% Entitlement for each VM machine ("server3" & "server4") . Can i able to do
this ? If Yes then can you give me the breakdown of the CPU distrubtion , If not
also please give the reason in detail?



Now My question


I think in this case we can create a server 3 and server 4 vm with 5vcpu but if server 1 and server 2 are already running then the server 3 and server 4 Vm will not start.

Because My understanding is that if server 1 and server 2 are already running and server 3 and 4 are started , then the later will get 30% entitlement each from leftover 4 cores , but for 5th vcpu it will not get the 30% as already 80% entitlement has been used by server 1 and 2 from the other 8 cores.

So i think server 3 and 4 VM's will not start. Please clarify whether my understanding is correct or please correct it.

[Torsten.]

See my comment on the same thread:

http://h30499.www3.hp.com/t5/Virtualization/Query-on-vCPU-distribution/m-p/4639703#M2202

[raiden]

Hello Torsten,

Appreciate your answer but Still I am confused over this question. I dont want answer in term of rx6600.

I just want a clarification on the above mentioned situation . (for now consider it to be any server) So please answer in terms of the above query .

I just want to know whether in the above situation i can create a 30% entitlement server 3 and server 4 VM and if i can create whether they will boot when server 1 and server 2 vm are already running.

Because the only question that bothers me is from where the server 3 nd 4 will get the 30% entitlement for 5th vcpu if 80% is already been used by the server 1 and 2 vm from the 8 cores.

[raiden]

I hope my question is not confusing you people . If it is please let me know . I will try my best to simplify it.

[Torsten.]

IIRC you never entitle "per core". The host has 100% to give with all of the CPUs, all the guests have to share the 100%. Start with a low entitlement, if needed, set a maximum. If a guest needs more power and there is more power available, the guest gets more power.

[Torsten.]

The problem with the calculation in the initial post is, there not 12, but only 8 physical cores in this type of server (maximum). So any calculation based on 12 cores must be wrong.

[Torsten.]

Finally, if you go back to the other thread (see link above) you'll find a calculation. The problem is, the result is wrong if you calculate with more cores than you have - there is nothing available while you think it is.

[Emil Velez]

A one processor system can create many VMs and run multiple VMs concurrently with VCPU.

A 4 processor system can create many VMs and run them concurrently with Vms with 1,2,3 or 4 CPUs
Create a VM with 4 VCPUs
Create a VM with 3 VCPUs
Create a VM with 1 VCPU
and I can run them concurrently if I have enough memory with only 4 real CPUs. I define a real cpu as one that top sees. WIth dual and quad core systems a cpu is a core. (not hyperthreading)

[Torsten.]

Emil, the question is about the entitlement values.

[likid0 · Accepted solution]

raiden you are write, the 5th virtual cpu only has 20% free, and needs 30% free to startup:

1rst example:

4 core Host
2 VM's with 4 vcpus each:

VM1 with 90% entitlement and running
VM2 with 10% entitlement

VM2 starts ok.

2nd example:

4 vore Host
VM1 with 4vcpus
VM2 with 1vcpus

VM1 with 90% entitlement and running
VM2 with 15% entitlement

VM2 won't start, because none of the cpus have 15% free.

That's why you don't have to use big entitlements, always assign small values and if you need, set the maximum.