Clarification about type 1 hypervisors and local storage

[RemoteDev]

Greetings. Long time browser, first time poster. I'm hoping the smart group here can help clarify some things for me per my specific setup.

My previous virtual computing experience resolves around two areas, mostly. First, an administrator on an all-Sun lab running multiple Zones on T2/T5 servers. Secondly, using VMware Workstation and Virtual PC on my desktop for remote access to client sites. Simple stuff.

However, apart from all that, as I've moved on to a new employer, I'd like to setup an ESXi-esque environment at home. The main purpose is development and testing. Ideally, I'll have 3-4 VMs running Windows and Linux, all relatively light-weight (i.e., no SQL servers, mail servers, video encoding, etc.) using ~250GB of disk space each.

After reading various sources such as here, Google and documentation, I'm confused about the storage option I'd like to use. I envision that the setup would sort of mimic a desktop VM install where all the VM's are using VMDK files stored locally, on disks directly attached to my motherboard. I would rather not run these via something like VMware Workstation as I'd look for them to be running indefinitely.

I see the big hangup is the availability, or lack of, VT-D. I think I understand the importance of that to efficiently and effectively pass through a controller (or otherwise) directly to the VM. However, I wonder if that applies to my circumstance. I don't think so but it's been confusing finding any examples that would apply to my case.

Here is what I would like to do:

1) Purchase components, which may or may not be a motherboard with VT-D capability, with a 2TB SATA disk local to the machine
2) Install ESXi (or Xen, KVM, etc)
3) Provision the disk to store the VMDK's for the various VMs (still fuzzy on the exact steps)
4) Install and run 3-4 VMs to run at the same time, all the time (potentially)

I understand this is probably outside the scope of what most people use ESXi for. However, here is what I'm NOT doing:

*) No production or customer use
*) No need to move VMs, ever, there will only be one machine
*) No need to extract away the storage structure or store data elsewhere (SAN, NAS, etc)

Note, any data I will care about will be backed up over the network.

With all that stated, here are my questions:

A) Do you think ESXi/Xen/etc. will suffice for my environment?
B) Is there documentation to configure such a setup?
C) Is there anything else I need to consider?
D) Is there another product or process I need to consider?
E) If I have multiple disks on the motherboard, are the all presented to the hypervisor as one or multiple volumes?

Thanks for the help. If I've overlooked something please point me to where I should look for answers.

 

[J-Will]

ESXi will work for your environment, and there is plenty of documentation out there on how to configure the host environment.

The biggest thing you need to consider is your hardware. It has to be compatable with the host environment, such as ESXi really likes Intel NICs and does not play well with most others. In order to get an idea of what harware you need, you should plan out your VMs in order to know their requirements piled on top of each other, and stacked on top of your host OS requirements. This is can be difficult. You'll want to provide plenty of RAM and CPU power.

Harddrives are presented as individual disks, unless of course a RAID solution is used. Usually, the host OS is kept on a dedicated harddrive and the VMs share some sort of RAID array that provides speed and redundency.

 

[J-Will]

VMs essentially become a file(s) on the hdd. The amount of space each VM uses on a hdd depends on the amount allocated to it, the amount within the allocation that it is using (Read: a VM given 100gb does not mean with will take up 100gb until fully used, and then it may exceed 100gb with added VM/ host config files), as well as the host config files for that VM. With multiple VMs running you may want to consider a RAID array or multiple disks. What you do on the VMs does not necessarily matter since all the VM OSs will be fighting over hdd use, but the problem grows when you are using VMs that are more mission critical and/ or need frequent hdd reads/ writes.

 

[RemoteDev]

J-Will,

Thanks for the good responses. So, based on your response ESXi would basically act as a really fancy VMware Workstation - just closer to the bare metal? That's what I'm going for.

These are not mission critical VMs - not even production, really. They will mainly be running an IDE as well as performing some long-running tests and deployments, replicating a specific environment. Only one will have any real HDD usage, and I assume that would be pretty minimal. I could probably do it via Workstation or Virtualbox if this route turned out to be overly complicated.

Per the hardware, I was going to look at a Supermicro or Asus motherboard with 8 GB of RAM, 3 SATA disks and an E3 Xeon - it's not too bad price-wise and can always be re-purposed later if something else comes up.

These boards usually have some sort of Intel NICs, though getting a separate NIC isn't too big of a deal. I'm assuming the NIC can be bridged and shared for all the VMs - I really don't need anything complicated for this environment.

Thanks for the feedback.

 

[J-Will]

ESXi is an operating system, so it sits directly on top of the bare metal hardware. From the hypervisor gives the VMs access to the hardware as needed.

I think your hardware plan is good. Its much better to get 'more hardware' that what you need now, then find out later you should have. Even for home/ just for fun use.

Most server class boards do have Intel NICs yes. And while you can do NIC teaming, you should always have at least two NICs on a server with VMs. One for exclusive use by the host- this allows you to always be able to remote access the host. The second NIC for use in the virtual network for the VMs. Of course if you get add on NICs, VMs with heavy network traffic loads could have their own NIC.

 

[RemoteDev]

Perfect, thanks again.

 

[defuseme2k]

ESXi is an operating system, so it sits directly on top of the bare metal hardware.

Don't call it an operating system. Though it technically sits in place of one, I wouldn't call it an OS. Yeah I'm picky