67 replies to this topic

[ikon]

You have summarized it well. The keyboard mouse stuff is handled when you RDP into the machine. ESXi uses proprietary drivers and Hyper-V uses default MS drivers. As for the link, I am not familiar with it but I would not use a software solution for something important.

Agreed, nice summary, except for the tiny font size.

Also, given that that hardware device is so cheap it hardly seems worthwhile to bother with a software solution.

[jmwills]

Whadda you say?

[KrisseZ]

Has someone pointed out that VMwares pass trough capabilities are hardware bound? That's why I decided to drop my ESXi project cause I couldn't point the HDDs directly to a VM.

[ikon]

Whadda you say?

Speak up, I can't hear you.

[Server Grunt]

Haha!!
Sorry for the small font - I do not know what happened :-)
Anyhow - I have decided to go with Hyper-V and i am currently trawling the Internet after all info I can get my hands on.

As I write this i am burning the ISOs for server 2008 R2, WHS2011 as well as for W8

I am really looking forward to exploring this new area!

As always - a massive shout out of gratitude to all of you guys!!!

/Grunt

[jmwills]

No need to burn ISO images. Install Virtual Clone Drive which is a virtual CD/DVD drive and you can mount the iso images on the fly when needed,

[Server Grunt]

I know, but like the physical discs when doing my installs first time around.

[jmwills]

I can't tell you the last time I burned a disc.

[ikon]

I can't tell you the last time I burned a disc.

For bare metal I usually burn an empty USB flash drive or, better yet, just install the ISO to a YUMI USB flash drive.

[Server Grunt]

Installation related question
Setting up the RAID for the storage disks and I can choose to set sector size between 512B and 4k
Server 2008 R2, WHSwhs 2011 and w7 supports 4k, but as I understand it Hyper-v does not natively support 4k.
Can I still use 4k sectors in the storage RAID or should I go 512B in a hyper-v environment even though both host and virtual machines support 4k?

/Grunt

[ikon]

Hmmm, I'm thinking, if you're not passing any hard disks through to a VM, the drives in the RAID will be virtualized and so the guest OS won't know, but I could be completely out to lunch.

[no-control]

If you're going to pass through the array to one of the operating systems that supports it then use 4k. If it's for VM use 512.

[darkside34]

For exaple as I understand it Hyper-V enables dynamic memory allocation, something that ESXI does not.
/grunt

ESXi Does not need Dynamic memory allocation. ESXi runs each virtual machine as its own process, meaning each virtual machine will only occupy as much of the hosts memory as It actually physically uses via the standard Linux memory scheduler. This means you can over-allocate a server without consequence (although if you run out some of your hosts will thrash). I have attempted to use dynamic memory allocation in test environments at work for clients, but it usually leads to bluescreens (Likely the fault of whoever wrote PERC storage controller drivers, which enjoy eating host memory without letting the OS know).

Hyper-v on the other hand, operates much like an IBM LPAR. The host system itself is actually a virtual machine, and the virtual machines are all run in parallel to that. Which means you can only statically assign memory. All dynamic memory allocation does is dynamically change the static assignment .

Also, why does people think running off a thumb drive is an advantage. Cool, yes; advantage, I don't see it.

Bare-metal ESXi is designed to load itself into the memory of the system. ESXi only performs I/O operations when you make settings changes or writes logs, literally everything else is stored in memory. This makes the lives of sysadmins in server farms a lot easier since most hypervisor-purposed servers have an internal USB port. It also allows to run the servers completely disk-less with all RAID/SATA functions disabled.

And if one ever fails (not that I have ever seen one fail) Just DD the ESXi image onto a new flash drive, replace, and go.

Edited by darkside34, 16 April 2012 - 01:41 PM.

[jmwills]

ESXi is great is your board supports VT-d which will allow you to pass through RAID arrays and drives larger that 2TB. Or at least, that is my understanding of it.

[darkside34]

ESXi is great is your board supports VT-d which will allow you to pass through RAID arrays and drives larger that 2TB. Or at least, that is my understanding of it.

VT-d has nothing to do with how large of a disc drive you can use. Older versions of vmware didn't support GPT booting, meaning you couldn't boot an OS on a 2.5tb or larger partition. VT-d may have a role to play in the size of a disc you can PASS THROUGH, but I am not sure as that is a feature I never use.

I still don't understand why people insist on passing through drives direct to hosts, there are easy ways to get around the partition limit without totally compromising host performance and making a very inconsistent and convoluted setup that doesn't allow you to share your storage with other virtual machines. You could always pick a bigger block size. Or give your file server multiple 500gb virtual discs and use LVM to span all of them. Of course windows-only users are kind of screwed in that sense.

[jmwills]

That's where the learning curve comes in for some of us. There are cases where I want to pass thru drives to dedicated hosts.

[ikon]

And most people here are wanting to load Windows Server of some flavour so some of these 'tricks' aren't going to work.

[darkside34]

And most people here are wanting to load Windows Server of some flavour so some of these 'tricks' aren't going to work.

Yup. Well I would argue that windows hyper-v probably has much less of a learning curve than ESXi, but in the long run ESXi will definitely give you more mileage.


Of course, I drank the open-source kool-aide and use KVM as a result of such. There is always proxmox (KVM on easy mode) It is easy, the interface isn't that nice, but it doesn't have as many limitations as ESXi and bad performance of hyper-v.

[ImTheTypeOfGuy]

VT-d has nothing to do with how large of a disc drive you can use. Older versions of vmware didn't support GPT booting, meaning you couldn't boot an OS on a 2.5tb or larger partition. VT-d may have a role to play in the size of a disc you can PASS THROUGH, but I am not sure as that is a feature I never use.

I still don't understand why people insist on passing through drives direct to hosts, there are easy ways to get around the partition limit without totally compromising host performance and making a very inconsistent and convoluted setup that doesn't allow you to share your storage with other virtual machines. You could always pick a bigger block size. Or give your file server multiple 500gb virtual discs and use LVM to span all of them. Of course windows-only users are kind of screwed in that sense.

To my knowledge drives are not passed through to hosts as the host sees all drives directly. Drives are passed through to VM's!

[ikon]

To my knowledge drives are not passed through to hosts as the host sees all drives directly. Drives are passed through to VM's!

Ah. Interesting. I took his use of hosts to mean the VMs. Now I'm wondering....