Slots size and amount of slots availables.

[projectja]

We have  2 ESXI 5.0 with exactly the same hardware

18 virutal machine split between two nodes. HA and DRS configured.

Now, we are trying to clear up and understand what is the slot size and how Vcenter calculates slots values so we check whether those ones match up with the ones we get by ourself with manual calculations.

As I said, we are trying to understand concepts, however values shown in Vcenter does not match up with our calculations.

The resources allocation values shown in Esx1 and Esx2 -resource allocation tab-  for Host-1 and Host-2 are shown in the image attached.

Our Vmware infraestructure:

- No reservations for cpu at virtual machine level.

- No reservations for memory at virtual machine level.

- A resource pool -for testing- with memory and cpu reservation established but theorically these values are not taken in account for these calculations.

Questions:

- Vmware architecture theory states that 32 mhz is taken when not CPU reservations are established:

*When we are calculating, should we multiply 32 mhz by the amount of VCPUs ? (Example: vm1 = 32 mhz x 8 cpus , mv2 = 32 mhz x 1 cpu and so on...)  -obviosly when when not reservations are established at vm level-

- Overhead memory, -supposing not reservation are established-:

*When we are making slots size calculations,  we should take the overhead memory parameter per each machine in orther to take the greater value of memory. Is that right?

To be honest we are not able to get the same values as Vcenter is showing.

Attached the  pictures of our infraestucture in lab.

We have the book of Ducan, it is wondelful. Also we have read many articles but it is not clear for us.

Please, any guidance about Vcenter is calculating, will be welcome.

[a_p_]

I assume I don't need to explain what slots are since you already have the HA book!?

Regarding the number of slots, the calculation is quite easy. What you do is to divide the total CPU by the value for the slot size (30,387 /32 = 949 Slots) and then you divide the total memory by the value for the slot size (125,495 / 299 = 419 Slots). From these two values, the lower one counts.

André

[projectja]

Thanks very much. You've helped me with that simple calculations. Sorry forthe post, yestarday I was very, very tired and I made some mistakes when writing the post, we have edited it.

It was not clear for us. We were not sure whether 32 mhz was a value per vcpu or by the contrary it is value per virtual machine as a whole, I mean, whether we had to multiply 32 mhz by each vcpu a virtual machine has. With regard to your explanataions, Therefore I understand the value is absolute value to the virtual machine without taking in account number of vcpus.

From here, we are going to create an excel containing calculations, overheaed memory inputs per virtual machine, Cpu, etc

Obviosly, with resources reservations playing, the scenario will change, Probably if we have some servers with high reservations we should to choose Percentaje Adminission control.

For example in Duncan book, there is an scenario e with 8 hoist ,12,5 % and also another one with 12 hosts. I think, and the percentaje security level parameter will change,  Is that right?

In our case, with two host and percentage admision control policy, supposing two indentifcal server, the percentaje would be 50% ?

Thanks agen for your feedbackd

[a_p_]

Yes, with only 2 hosts the percentage setting (~100% divided by the # of hosts) may be more appropriate then the number of host failures. Only the "Host failures tolerated" setting uses the slot number calculation for Admission control. Although this setting might be the most secure in regards of ensuring enough resources to restart VM's in case of a host failure, it may not be the desired setting if VM based reservations are used.

André

[emma234]

Hi projectja

Welcome to the communites

On single line statment I will say yes  and your calculation are right regarding below.

In our case, with two host and percentage admision control policy, supposing two indentifcal server, the percentaje would be 50% ?

"He Conquers, Who Conquers Himself".