Hi,
I do not have any background on VMware technologies, but we are using a dynamic cloud installation based on vCloud Director at the company where I work.
The group which operates the server advertises the service offering with a "pay per use model".
This is true for CPU resources, as we pay only what we use during the month. But in my opinion this does not hold for RAM usage.
We have Linux VMs which allocate 4GB of "physical" RAM and our monthly bills show that the RAM usage is up to 100% for the whole month. But when you look at memory tools of the VM you can see that most of the time 50% or even less is used by the guest OS processes.
That means from a user point of view the RAM price per month should be half of what we are charged right now. Asking the operators about the "issue", we always get the response that RAM usage is not dynamic and we need to pay for the 100% usage from the host server's point of view.
Could someone share some light on this topic? Is there really no way to make RAM usage as dynamic as CPU usage?
Thanks,
Michael
Evening,
As a service provider I'll let you into the secret:
It depends
vcloud supports three allocation models by default. Your provide might use others
-Pay as you use. - pay as you use
-Allocation - you get allocated resources but can burst higher
-Reservation - you get allocated resources and that's all
Chargeback depends on model (in out of the box chargeback... your vendor might have something custom)
Pay as you use - charged based on used resources
Allocation - Pay based upon allocation then additional for burst
Reservation - Pay based upon allocate
So in the last two models there really is no way to charge except full allocation (in other words 100% cpu and RAM)
Based upon you statement I think you have something along the lines of pay as you use mixed with a vendor pricing scheme.
In my experience the resource most used in ESXi clusters in RAM. It's really what I buy in resource clusters... I always need more RAM. I never need more CPU. In addition allocation of CPU in vcloud is done via Ghz's not CPU's. RAM is allocated by GB. So put those together and I think you providers is betting that they will always need RAM so they charge based on allocation not usage. While CPU they have tons so they charge by usage.
Really this is all just my thoughts...What can you charge on:
-Ghz of CPU
-GB of RAM
-GB of storage
How can you charge:
-Used
-Allocated
So end result... they can charge by used they choose not too..
Hope it helps,
J
Evening,
As a service provider I'll let you into the secret:
It depends
vcloud supports three allocation models by default. Your provide might use others
-Pay as you use. - pay as you use
-Allocation - you get allocated resources but can burst higher
-Reservation - you get allocated resources and that's all
Chargeback depends on model (in out of the box chargeback... your vendor might have something custom)
Pay as you use - charged based on used resources
Allocation - Pay based upon allocation then additional for burst
Reservation - Pay based upon allocate
So in the last two models there really is no way to charge except full allocation (in other words 100% cpu and RAM)
Based upon you statement I think you have something along the lines of pay as you use mixed with a vendor pricing scheme.
In my experience the resource most used in ESXi clusters in RAM. It's really what I buy in resource clusters... I always need more RAM. I never need more CPU. In addition allocation of CPU in vcloud is done via Ghz's not CPU's. RAM is allocated by GB. So put those together and I think you providers is betting that they will always need RAM so they charge based on allocation not usage. While CPU they have tons so they charge by usage.
Really this is all just my thoughts...What can you charge on:
-Ghz of CPU
-GB of RAM
-GB of storage
How can you charge:
-Used
-Allocated
So end result... they can charge by used they choose not too..
Hope it helps,
J
Thanks you very much for the information.
I will have a talk with our providers now
Thanks,
Michael