vCloud director vCPU speed setting(VDC allocation model) and performance metrics(Cassandra DB)

[ManivelR]

Hi Guys,

I have two doubts.Could you please suggest?

1) Cassandra DB:- I completed the Cassandra DB setup this in lab before moving in to production.

In lab,1 VCD VM,1 VCD postgress DB VM and 1 Cassandra DB VM. By default,it was 8 parameters in monitoring(after integrating Cassandra DB with VCD cell).

I just added network parameters in configuration file.If i want to add any more parameters in future,should i need to edit the same configuration file ? or should i use the different one(new one as “metrics.groovy_1” ? Please suggest.

2) Its regarding vCPU SPEED setting in allocation model( in VDC).I have a slight confusion about GHZ vs vCPU.

My ESXi server CPU model and speed is mentioned below

Im using  "Allocation pool". Example VDC is given below.By default the vCPU speed is 1 GHZ and the i have given the CPU allocation as 12 GHZ.

Doubt:- If i want to provide 12 vCPU to the same VDC,then i need to change the vCPU speed as 2.6 GHZ(because my ESXi speed is 2.6 GHZ) and then i need to set the CPU allocation as 31.2 GHZ (vCPU speed 2.6 GHZ * 12 VCPU = 31.2 GHZ).

Is this right or can i set 31.2 GHZ(CPU allocation) with VCPU speed as 1 GHZ(default one?

Thanks,

Manivel R

[sk84]

1) If you want to add more counters, the configuration file must only contain the new counters. Otherwise vCD will complain when importing the file that the counters already exist.

2) If you have an allocation pool with 12 GHz CPU allocation and a CPU speed of 1 GHz, the customer can use 12 vCPUs (12 * 1 GHz). Technically it will set a CPU limit of 1000 MHz in the ressource settings on every VM. So it doesn't matter what physical cpu speed you have.
But the real reservation on vSphere level depends on the setting "CPU ressources guaranteed" and if it's an elastic allocation pool or not. If you only guarantee 10% and have elastic allocation pool enabled (default setting), vSphere will set a reservation of 1.2 GHz ((12 * 1 GHz) / 10) on the ressource pool of this orgVDC. In short: These values only serve to control the CPU overcommitment.
Btw, if you set the guaranteed CPU ressources in the allocation pool model to 100% you have basically a reservation pool (all configured ressources are guaranteed).

However, I would consider that nowadays many applications require more than 1 GHz and therefore customers will need more vCPUs. And the more vCPUs a VM has, the higher the CPU ready values can be, resulting in poorer performance. Because the ESXi CPU scheduler must find a free slot for _all_ configured vCPUs of a VM (independently from the usage). And the more configured vCPUs a VM has, the less likely ESXi can assign a scheduling slot in time. It is therefore usually more advisable to configure as few vCPUs as necessary. But that also depends on your resource management. If your ESXi servers have large enough CPUs and the overbooking is low, it should be no problem.

[ManivelR]

Thanks so much Sebastian for your inputs. I made allocation pool as elastic.Lets say, i created a demo7 VDC with below settings. 12 GHZ = 12 vCPU.

I was not created any VMs under this Demo7 VDC.If i want to create any new VMs,lets say 6 VMs * 2 vCPU,then it will allow us to create as the allocation pool is elastic(eventhough the CPU resources guaranteed as 10 %).When Im touching the CPU usage at 1.20 GHZ(lets say 2 VMs are on now),then i cannot power on any other VMs(rest 4 VMs).

Am i right?The same applies to memory as well.

In VDC:-

in vCenter server resource pool settings.

Thank you,

Manivel R

[sk84 · Accepted solution]

Elastic Allocation Pools only plays a role for you as a provider, where and how the resources are allocated (on VM level or RP level for example).

And also the CPU guarantee value is only for you as provider to overcommit the CPU resources. If you guarantee 10%, in vSphere only 10% of the used CPU resources will be set as reservation. But for the customer it looks like 100%. So, if the CPU ressources are completely used the customer can not power on any further VMs.

As an example:

1 GHz CPU speed, allocation pool with 12 GHz and 10% CPU guarantee. The customer configures VMs with 12 vCPUs.

In the vCD GUI the customer now sees that 100% of the CPU resources are in use. But in vSphere only 1.2 GHz are reserved for the customer. So you as a provider can overbook the CPU resources. That's the advantage of the allocation pool in contrast to the reservation pool model where all resources are bound to the customer.

[ManivelR]

Thanks so much Sebestain for your response.

[ManivelR]

Hi Sebestian,

Only one doubt about Cassandra DB. Im using VCSA 6.7.0 and there are totally 563 monitoring parameters available.

I need VM disk IOPS parameter and i could not see this in VCSA parameters(563 parameters).

Do you have any idea on this?

Thank you,

Manivel R

[ManivelR]

I got it Sebastian.I got the Cassandra database metric path(Metrics location from Yaml file.

Thank you,

Manivel R