Recommendation & Advice Please

[jitla1971]

Hi Everyone

Need some advice.

I going to create a Windows 2008 R2 Standard VM.

Client wants a C, D, E & F drive within windows.

Would you recommend create 1 vmdk and partition it via Windows or would it be better to create a separtate vmdk for each required partition.

Are there any best practice guidelines for Windows 2008 VM creation?

2nd question i have relates to vcpu and what is the best method to assign.

We have a HP DL360 G6, vi client is showing

CPU Cores: 4 CPU's x 2 Ghz

Processor Socket 1

Cores per Socket 4

Logical Processors 4

I get really confused with vcpu as ive not been able to find a simple explanation, as i cant get my head round this concept.

Is there a rule of thumb, when it comes to assigning a vm with more than 1 vcpu.

Does the vcpu correspond to the amount of cores?

If i assign a vm 2 vcpu would it use 50% of the available cores and will it require 2 cores to be available before it starts any cpu related tasks.

with current setup, what would you recommend for vcpu config.

Sorry to ramble on so much, its just confusing.

Jitla

[MarkStrong]

I would use separate VMDK files for D:, E: etc

In most cases try to use one vCPU.

Some light reading:

http://www.vmware.com/files/pdf/perf-vsphere-cpu_scheduler.pdf

VCP4 | MCITP | HP Master ASE | CCNA

[fusionit]

First in regards to your disk question. This is a somewhat loaded question as there are many variables which determine how you do this.

1. how large are the partitions?

2. is the VMFS store local, on a a SAN, or on a NAS?

3. What will be stored on these partitions? (files, databases, etc)

I don't know about best practice but I have always made it a rule of thumb not to go larger than 500GB virtual disks when using local storage and 256GB virtual disks when using SAN or NAS for production environments. But this is a general rule of thumb because it really depends on what's running on it.

simple file storage will run fine on larger volumes where Databases that get hit continuosly tend to run better on smaller partitions.

With regards to type of storage ... this is the age old equation more spindles = better performance. if you have a 500GB vhd running sql databases having it accros a raid 50 array with 26 10K disks is going to give better performance than a raid one 2 disk mirror on 7200 RPM disks.

As for VCPU's I don't realy know much about the "under the hood" metrics of how they map but I know for sure it is not a one to one ratio I ahve 25 VM's running on a host that has two dual core CPU's and many of the VM ar erunning more than one CPU if that helps.

[J1mbo]

I use seperate VMDKs most usually since these can be made larger in the future easily if needed.

For vCPU info, try this: http://blog.peacon.co.uk/understanding-the-vcpu/

http://blog.peacon.co.uk

Please award points to any useful answer.

Unofficial List of USB Passthrough Working Devices

[jitla1971]

Hi

Thank you for your response.

We are going to use the following partition sizes.

C: 80GB

D: 100GB

E: 300GB

F: 80GB

The vmdk files will be stored on the local (SAS disks non san or nas) datastore, which is in vmfs format.

Thank you all again for your help.

[ezzeldin72]

it is recommended to create one VMDK for each partition for better managability.

vCPU is not mapped one to one to logical CPU. you can schedule as much as you need vCPU to one Logical CPU. the key point is the workload. e.g. you can created 20 VMs on your servers (one vCPU for each vm) and all VMs can work fine on the 4 logical CPU you have as long as they do not need extra CPU cycle (if it happend, some VMs will have to wait till the other finish)

Ezzeldin, VCI,VCP3,VCP4,VSP,VTSP

Kindly note, if your problem or question has been resolved, please mark the thread as answered and award points.

[ezzeldin72]

it is recommended to create one VMDK for each partition for better manageability.

vCPU is not mapped one to one to logical CPU. you can schedule as much as you need vCPU to one Logical CPU. the key point is the workload. e.g. you can created 20 VMs on your servers (one vCPU for each vm) and all VMs can work fine on the 4 logical CPU you have as long as they do not need extra CPU cycle (if it happend, some VMs will have to wait till the other finish)

Ezzeldin, VCI,VCP3,VCP4,VSP,VTSP

Kindly note, if your problem or question has been resolved, please mark the thread as answered and award points.