Whether to virtualise our main servers

[chazzy2501]

Virtualise, it'll be the last upgrade you ever do. (maybe not) p2V your hardware servers onto a new virtual environment (you can schedule this) then fire 'em up change the network card settigns and they'll be up and running like nothing changed in minutes. Once virtualised you can pick em' apart with the security of snapshots etc. The stress relief of being able to role back a fluffed setting or restoring an entire copy of the server if something goes wrong is bliss. no more ASR disks!!!!

[Jollity]

Thank you again all for further useful advice.

 

Next up, work out how many CPU cores, Ram and disk space each of those individual VM's need. From this you can work out how meny CPU cores, Ram and hard disk your physical servers need. For a 100 computer school I think it'd be tough to justify more than 2, maybe 3 host servers.

How would you go about working out CPU, RAM and disk space requirements for a VM? Do you measure performance on existing servers or guestimate? How much overhead should be allowed for the OS on each VM (particularly how much disk space for system files)?

 

i would also recommend if you are going to get a machine at the other end of the building (like current) then I would Custom build a powerful desktop with SATA drives to replicate using Veeam to.

The example server spec is really helpful, glennda. Are you suggesting the powerful desktop could actually be used as a virtual host if the main server is down or is the idea just a quickly accessible copy of the virtual machines in case of disk failure?

[Jollity]

Virtualise, it'll be the last upgrade you ever do. (maybe not) p2V your hardware servers onto a new virtual environment (you can schedule this) then fire 'em up change the network card settigns and they'll be up and running like nothing changed in minutes. Once virtualised you can pick em' apart with the security of snapshots etc. The stress relief of being able to role back a fluffed setting or restoring an entire copy of the server if something goes wrong is bliss. no more ASR disks!!!!

Certainly sounds like a plan for the other servers, though probably as a phase 2 after upgrading the main ones. Do you have any issues with the drivers in doing this? Presumably the virtual hardware cannot perfectly reproduce the original?

[tmcd35]

How would you go about working out CPU, RAM and disk space requirements for a VM? Do you measure performance on existing servers or guestimate? How much overhead should be allowed for the OS on each VM (particularly how much disk space for system files)?

 

It's exactly the same as spec'ing physical servers to do the same tasks. What does the OS you are using recommend? What does the developer of the software you are using recommend?

 

If you are virtualising existing servers then yes look at current performance measures for those machines.

 

Also, if using Hyper-V, don't forget to spec an additional core and 2Gb for the host OS. (I'm not sure what VMWare asks for the host, probably 1core and around 512mb).

[cpjitservices]

Make use of VMWare Converter it's a free tool. You can convert any powered on machine to a VMware ESXI Host (Converter classes it as a Physical Machine).

[zag]

Virtualisation is a total no brainer, modern servers are totally bored with the load of a standard windows server, just add more ram and make them run 10 or 12. I've got a big environment, but I would virtualise in any environment just for the benefits of role separation and the ability to add a new server for any new roles that come along.

 

Veeam for backup is an absoloute marvel. Mine runs through about 6TB of servers every night in about 6 hours.

 

Totally agree with this!

[fiza]

Make use of VMWare Converter it's a free tool. You can convert any powered on machine to a VMware ESXI Host (Converter classes it as a Physical Machine).

 

You can do the same type of thing with Hyper-V using Disk2vhd

if you choose to use Hyper-V instead of VMWare

[Trojan]

Another +1 for virtualisation here. VMware essentials plus, 3 Hosts, SAN, Veeam B&R - Sorted.

 

Cannot recommend it enough, less physical space used, less power consumption, more efficient use of hardware, backup and DR covered, redundancy and HA built in. It's a marvel.

[Domino]

Certainly sounds like a plan for the other servers, though probably as a phase 2 after upgrading the main ones. Do you have any issues with the drivers in doing this? Presumably the virtual hardware cannot perfectly reproduce the original?

 

After a p2v I'd always recommend removing the redundant hardware devices (as the virtual hardware is *new devices* ) Removing old hardware after a P2V conversion - Virtualization Pro

 

However, where possible, I'd recommend building new boxes and moving over to them anyway

[Jollity]

I have been starting to do some reading and working out specifications. I am tending towards Hyper-V mainly as it seems, in the limited time available before bring it into use, that it would be easier for my colleagues and me to get our heads round, as Windows is what we are already used to, but I am not set on that.

 

I was wondering what specification people used for a minimal server 2012 guest computer. A DC, for example, presumably does not need much in the way of memory or other resources. The minimum RAM requirement of Server 2012 is supposedly 512MB, but would you actually stick at 2GB?

 

Similarly for hard disk, how much space do you give a windows partition? Or do dynamic disks make this fairly unimportant? It seems a pity to have to store 20GB+ of identical system files for each guest, but I suppose that is part of the price for the benefits of virtualisation.

 

Do people tend to remove the GUI from guest Windows installs?

[robjduk]

My VMWare DC's use 4GB each. I am moving to 2012R2 soon and imagine they would be fine with that too. As for partitions, my 2008 ones have 40GB as a HHD partition but 60GB seems to be the norm nowadays but would think you could still get away with 40 GB easily.

I only run one server core and it is great but only did it as a training exercise. If you are happy with server core then knock yourself out however if not then standard will be fine in a hyper V environment.

[seawolf]

Yes!!

 

Get 1 x this - HP ProLiant DL380p Gen8 E5-2690v2 2P 32GB-R P420i/2GB FBWC 750W RPS Server(709943-421)

 

add another 32GB ram into it. Then stick 8 x 450 or 600gb drives in it. If you are going local storage ensure you get a 2gb FBWC raid card. also don't install ESXI to USB stick if going the VMware root, install to a Good Quality SD Card (these servers have a slot directly on Mobo)

 

i would also recommend if you are going to get a machine at the other end of the building (like current) then I would Custom build a powerful desktop with SATA drives to replicate using Veeam to.

 

Note: a replication is not a backup, you need to ensure you backup also to Nas or similar.

 

Whoa...never use just one server in a virtualised environment. That truly is putting all of your eggs in one basket. You ALWAYS need at least two servers To serve as VM hosts. And, both should be able to run all of your VMs in the event of a server failure.

 

Veeam backup is a great recommendation, but you want your backup server to have lots of CPU grunt, at least 12GB RAM, fast storage and at least two NICs that can be trunked. Otherwise, your backups run slow, restores run slower, and you would barely be able to run more than one VM in an instant restore scenario.

[seawolf]

I have been starting to do some reading and working out specifications. I am tending towards Hyper-V mainly as it seems, in the limited time available before bring it into use, that it would be easier for my colleagues and me to get our heads round, as Windows is what we are already used to, but I am not set on that.

 

I was wondering what specification people used for a minimal server 2012 guest computer. A DC, for example, presumably does not need much in the way of memory or other resources. The minimum RAM requirement of Server 2012 is supposedly 512MB, but would you actually stick at 2GB?

 

Similarly for hard disk, how much space do you give a windows partition? Or do dynamic disks make this fairly unimportant? It seems a pity to have to store 20GB+ of identical system files for each guest, but I suppose that is part of the price for the benefits of virtualisation.

 

Do people tend to remove the GUI from guest Windows installs?

 

If you primarily or only run Windows servers, then Hyper-V is the clear way forward for you. If you do use Hyper-V, make sure you use Windows Server 2012 R2 with the latest updates for your Hyper-V hosts.

 

Here are my recommendations based on 7 years of running a virtualised infrastructure:

 

1. ALWAYS use at least two hosts for your VM environment. Using only one is foolhardy, more than two is unnecessary for most and just more expense and more to manage.

 

2. ALWAYS overspec your VM servers by at least 30%, especially for the CPUs which are much harder and more expensive to upgrade. The minimum RAM you should use is 1.5GB for Linux servers, 4GB for standard Windows servers (file, web, general purpose), 6GB for DCs and 8GB for Database servers. Always use at least 2 vCPU cores per Windows VM and 4 vCPUs for DB servers. Use a minimum of two NIC ports for every vSwitch (LAN, DMZ, SAN) and 4 or more are better for the SAN network (if you use a SAN). A bare minimum for most environments is two servers each configured with at least 2 x quad core Xeon CPUs (1 x 8 core or 2 x 8 core even better), 32GB of RAM and 3 x dual port NICs (6 total).

 

3. ONLY use a SAN if you know what you are doing. An improperly configured or managed SAN will give you nightmares and you are better off using DAS. Don't try to use an inexpensive NAS as your VM storage. Bad mistake. If you are spending less than $8,000 for your SAN - it isn't good enough to use for your primary VM storage. The lowest end I would go is with a mid-tier TruNAS server, a Drobo 1200i, or similar. Better yet, build a SAM-SD or buy a Nimble storage SAN or Oracle 7120/7320 (just discontinued might be able to find at a deep discount). Otherwise, just stick with DAS with good fast 10,000 or 15,000 RPM SAS disks.

 

4. Go with servers that can support 10GbE NICs. Even if you won't use 10GbE right now, it's good to have the option in the next 5 years. Remember that with a VM host ALL of your VM servers are sharing the bandwidth on the host rather than it being distributed across multiple physical servers each with their own independent NIC. Many often fail to realise this little fact. 10GbE really comes into its own with VM hosts and SAN storage.

 

5. BACKUP, BACKUP, BACKUP! Ensure you have a good backup server and backup storage. Don't skimp on this aspect of your environment. Use a server with at least a quad core CPU (i7 or Xeon), 12GB RAM, and two NIC ports. Ensure your backup storage uses enterprise rated disks such as the WD SE or RE series. A good option for backup storage is a Drobo B800i with 3 or 4TB WD RE drives. I recommend Veeam backup for your backing up your VMs. The newest version also supports secondary backup to tape or a secondary backup storage. I personally recommend building a cheap Micro-server running FreeNAS located in another building to copy your backups to just in case. I built one with 7.8TB storage for $1,600. Another option is an iOsafe fire proof backup drive connected to your backup server via USB 3.0. BackupAssist is also a great option particularly for backing up the Hyper-V hosts themselves (bare metal) and critical data backups.

[tmcd35]

Whoa...never use just one server in a virtualised environment. That truly is putting all of your eggs in one basket. You ALWAYS need at least two servers To serve as VM hosts. And, both should be able to run all of your VMs in the event of a server failure.

 

This I agree with 100%...

 

...This I don't (100%)...

 

 

 

2. ALWAYS overspec your VM servers by at least 30%, especially for the CPUs which are much harder and more expensive to upgrade. The minimum RAM you should use is 1.5GB for Linux servers, 4GB for standard Windows servers (file, web, general purpose), 6GB for DCs and 8GB for Database servers. Always use at least 2 vCPU cores per Windows VM and 4 vCPUs for DB servers. Use a minimum of two NIC ports for every vSwitch (LAN, DMZ, SAN) and 4 or more are better for the SAN network (if you use a SAN). A bare minimum for most environments is two servers each configured with at least 2 x quad core Xeon CPUs (1 x 8 core or 2 x 8 core even better), 32GB of RAM and 3 x dual port NICs (6 total).

 

Most of this is way, way over spec'd. Remember it's a 100 seat environment. You need to do a proper cost/benefit analysis. Start buy describing the servers you need, add a bit (30% sounds good, I'd agree with that) for future expansion, then double itif running two server (add a third for three servers, etc) so if 1 server goes down the remander can take up the strain.

 

I've never needed more than 3Gb in a DC and this is a 450 seat environement - DC2 currently sitting at 1.5Gb. DC3 using a bit more at 2.7Gb out of 12Gb (physical server, old VM host). 6Gb would be over kill, and 1 core is more than fine. I have VM's running on 1 core and 1Gb and others up at 4 core and 12Gb Ram. It's about matching the work load requirements. The best place to start is to look at the spec's and performance data from your existing machines.

 

Like wise, while it might be good to have 2xNIC's for each network. Your infrastructure needs to support it. More NIC's = more CPU/RAM requirements for the host OS to support them. Do your switches support 802.11ad? (that said 2012 now includes some very cool NIC bonding features).

 

3. ONLY use a SAN if you know what you are doing. An improperly configured or managed SAN will give you nightmares and you are better off using DAS. Don't try to use an inexpensive NAS as your VM storage. Bad mistake. If you are spending less than $8,000 for your SAN - it isn't good enough to use for your primary VM storage. The lowest end I would go is with a mid-tier TruNAS server, a Drobo 1200i, or similar. Better yet, build a SAM-SD or buy a Nimble storage SAN or Oracle 7120/7320 (just discontinued might be able to find at a deep discount). Otherwise, just stick with DAS with good fast 10,000 or 15,000 RPM SAS disks.

 

While I agree with the sentiment, the size of the environment says different. Personally I'd stick with local storage in each host server, but I think a couple of grand could buy a suitable NAS that support iSCSI with good enough speed/reliability for that environment. SAN's are rediculously expensive and I very much doubt is something you should be looking at.

 

4. Go with servers that can support 10GbE NICs. Even if you won't use 10GbE right now, it's good to have the option in the next 5 years. Remember that with a VM host ALL of your VM servers are sharing the bandwidth on the host rather than it being distributed across multiple physical servers each with their own independent NIC. Many often fail to realise this little fact. 10GbE really comes into its own with VM hosts and SAN storage.

 

All servers support 10GbE - it's called PCIe. Seriously, don't buy this now unless you need it. Buy it when you need it, when prices come down.

 

5. BACKUP, BACKUP, BACKUP! A good option for backup storage is a Drobo B800i with 3 or 4TB WD RE drives.

 

Pretty much this! I turn all my servers off once every term and backup to a 8Tb QNAP NAS. My entire virtual environment (nearly 30 servers + user data) fits inside less than 4Tb. Taken off site. If the school burned down we can get everything back very quickly and easily.

 

 

I suppose the point I'm trying to make is there is a danger of over spec'ing and spending too much (or getting a project mothballed on cost), as much as there's a danger in under spec'ing.

 

Other thing to bear in mind - Hyper-V pretty much requires atleast 1 physical domain controller (I'm aware 2012 R2 has "measures" to get around this particular chicken and egg, but...)

Edited February 3, 2014 by tmcd35

[sparkeh]

Whoa...never use just one server in a virtualised environment. That truly is putting all of your eggs in one basket. You ALWAYS need at least two servers To serve as VM hosts. And, both should be able to run all of your VMs in the event of a server failure.

Great sentiment but not always possible if you don't have the resources to achieve this (ie money available in a primary school!).

 

I have x1 host with 11Gb of Ram hosting Veeam backup and x2 Vm DCs on local storage, 1 running as a file and printer server, the other running SCCM, Sophos, Spiceworks and other stuff I can't think of right now. Everything (including Veeam backups) run very smoothly.

 

Not exactly as I would run things given the more money but it works well and I would rather choose this setup over physical servers.

[localzuk]

Everyone keeps talking of SANs for our environments, and I just don't think we need them still. I can't justify spending that much money on a SAN when I can get a DL380 with 25 disk bays for far less, and can populate it how I wish. Stick Windows 2012 on it and you've got a great environment for shared storage for Hyper-V.

 

In fact, that's what we have here.

 

So, you really don't have to splash out on expensive storage systems to have shared storage. For a small environment, just find a server with enough drive bays and spec the disks you want.

 

With Windows 2012, you can even have 2 such servers and specify a fileserver cluster which the Hyper-V cluster uses, so you have both Hyper-V and storage is redundant.

Edited February 3, 2014 by localzuk

[seawolf]

Great sentiment but not always possible if you don't have the resources to achieve this (ie money available in a primary school!).

 

I have x1 host with 11Gb of Ram hosting Veeam backup and x2 Vm DCs on local storage, 1 running as a file and printer server, the other running SCCM, Sophos, Spiceworks and other stuff I can't think of right now. Everything (including Veeam backups) run very smoothly.

 

Not exactly as I would run things given the more money but it works well and I would rather choose this setup over physical servers.

 

I strongly disagree. I think you would be better served by 3-4 less expensive physical servers with good warranties and vendor support and a good backup solution.

 

If you're SOLE server ever goes down (highly likely at some point), how long will your ENTIRE environment be down while you obtain a replacement server? Whereas, the likelihood of having multiple physical servers fail simultaneously is extremely unlikely. And, even in such cirucumstances only part of your environment would be offline. A good (and inexpensive) backup solution like BackupAssist can perform a full backup (bare metal) of physical servers for rapid restore.

 

Virtualisation is not the answer in every situation.

[seawolf]

Everyone keeps talking of SANs for our environments, and I just don't think we need them still. I can't justify spending that much money on a SAN when I can get a DL380 with 25 disk bays for far less, and can populate it how I wish. Stick Windows 2012 on it and you've got a great environment for shared storage for Hyper-V.

 

In fact, that's what we have here.

 

So, you really don't have to splash out on expensive storage systems to have shared storage. For a small environment, just find a server with enough drive bays and spec the disks you want.

 

I would agree with that sentiment, but disagree with the use of a single server for virtualisation. You need at least two or you can't afford to virtualise. Decent servers are relatively cheap when you only have the requirements of hosting one server on them. In situations of limited resources, it also usually means a small environment and one that is likely better off using a more traditional physical infrastructure.

[glennda]

I would agree with that sentiment, but disagree with the use of a single server for virtualisation. You need at least two or you can't afford to virtualise. Decent servers are relatively cheap when you only have the requirements of hosting one server on them. In situations of limited resources, it also usually means a small environment and one that is likely better off using a more traditional physical infrastructure.

 

Virtulisation provides many benefits in various ways (backup being the main) You don't always need 2 servers - I have lots of clients who have 1 primary ESXI host (normally Small Businesses under 10/15 users) then the server is replicated to an I5/I7 workstation with a couple 1TB HDD's in it (no raid). If the main server has a critical failure we can power on the second machine. This works well and normally saves 1/2k off the cost - yes they don't get the performance while failed over but it performs.

 

I have actually had to action this as a server had a raid failure meaning it needed to be bought online.

[zag]

Yeh we always have one spare server with Hyper-V installed. If the main machine fails, we just push out the backups from VEEAM to the new machine.

[sparkeh]

I strongly disagree. I think you would be better served by 3-4 less expensive physical servers with good warranties and vendor support and a good backup solution.

 

If you're SOLE server ever goes down (highly likely at some point), how long will your ENTIRE environment be down while you obtain a replacement server? Whereas, the likelihood of having multiple physical servers fail simultaneously is extremely unlikely. And, even in such cirucumstances only part of your environment would be offline. A good (and inexpensive) backup solution like BackupAssist can perform a full backup (bare metal) of physical servers for rapid restore.

 

Virtualisation is not the answer in every situation.

Why would I need a replacement server? Environment would be down as long as it takes HP to get here and fix it

If the worst happens we have a new server onsite next day and Veeam does its stuff.

And in the words of the Headteacher, in the event of a short downtime "we'll cope".

I'll trade the downtime we haven't had in the 6 years we have run this setup for the numerous benefits.

 

Virtualisation may not be the answer to every situation, but you don't understand every situation. I'm not worried.

Edited February 3, 2014 by sparkeh

[Jollity]

Thank you all for a lot of further useful information.

 

I am near certain we will go for local storage. Keeping things simple for one. Also not sure our current network infrastructure is fast or reliable enough to use for remote storage. Seemless automatic failover is I think not a requirement - recovery can I think wait for us to get another VM booted up.

 

DC2 currently sitting at 1.5Gb. DC3 using a bit more at 2.7Gb out of 12Gb (physical server, old VM host). 6Gb would be over kill, and 1 core is more than fine. I have VM's running on 1 core and 1Gb and others up at 4 core and 12Gb Ram. It's about matching the work load requirements. The best place to start is to look at the spec's and performance data from your existing machines.

The memory usage figures are really helpful. Can you say where you get the figures from? When I look into memory measurement, I keep running into complications such as software using more memory if more is available.

Edited February 3, 2014 by Jollity
typo

[tmcd35]

Can you say where you get the figures from? When I have looked into memory measurement, I keep running into complications such as software using more memory if more is available.

 

Crude - Just sat watching Task Manager for a few minutes. Soon get an idea of a servers load capacity. If software (SQL) is hogging all available RAM regardless, go back to the manufacturers recommended spec's, then take an educated guess as to whether or not you think the work load in your environment would need a little more.

 

IMHO, always run SQL server in a VM, and always run it on it's own. That way you can control and restrict it's memory creep and stop it from effecting any other roles.

[Trev_LCHS]

Like everyone else here, I can't recommend Virtualisation enough, its the dogs, it really is. We are using Server 2012 R2 HyperV. We have several things set up 2 hosts (nodes) that host everything on one San with cluster fail over, these hosts (nodes) are then backed up using Hyperv replica 3. This basically replicates your online servers to an offline replica which can be brought up in the event that you loose the main online nodes or the san. I back each node up to its own replica. I have one DPM backing up the data on those Servers in those Hosts (nodes) and another DPM backing up all the Virtual servers on the 2 replicas, this gets away from the issues that can occur when using DPM to back up clustered servers. so if my main san or nodes go down I just bring up the replicas, and once the main stuff is fixed replicate the changes back and bring them up, then carry on. The other beauty of Virtualisation is snapshots (now called checkpoints in server 2012 R2) ever wanted to roll your sims server back to pre update, easy snapshot it prior to upgrade trail it , SIM has stuffed the update up?, no probs just roll it back.

[seawolf]

This I agree with 100%...

 

...This I don't (100%)...

 

Most of this is way, way over spec'd. Remember it's a 100 seat environment. You need to do a proper cost/benefit analysis. Start buy describing the servers you need, add a bit (30% sounds good, I'd agree with that) for future expansion, then double itif running two server (add a third for three servers, etc) so if 1 server goes down the remander can take up the strain.

 

I've never needed more than 3Gb in a DC and this is a 450 seat environement - DC2 currently sitting at 1.5Gb. DC3 using a bit more at 2.7Gb out of 12Gb (physical server, old VM host). 6Gb would be over kill, and 1 core is more than fine. I have VM's running on 1 core and 1Gb and others up at 4 core and 12Gb Ram. It's about matching the work load requirements. The best place to start is to look at the spec's and performance data from your existing machines.

 

Like wise, while it might be good to have 2xNIC's for each network. Your infrastructure needs to support it. More NIC's = more CPU/RAM requirements for the host OS to support them. Do your switches support 802.11ad? (that said 2012 now includes some very cool NIC bonding features).

 

 

Yes, I do realise it is a 100 seat environment - today. However, in 12-24 months time, the school may throw 100 iPads into the mix or decide to roll out a fleet or laptops. And if that happens it is going to be difficult to go back to management and say "we need to buy new servers" when you've just put new ones in a few months ago. You always have to think about potential growth and not just the current environment.

 

As far as my recommendations on RAM, it does depend on your environment, but RAM is cheap. Are you really telling me that the average server should be configured with less RAM than the average laptop or desktop? 6GB is not over the top if you have a single DC. If you have two, drop it to 4GB, if you have three maybe you could get away with 3GB, but 4GB is minimum recommendation by Microsoft and VMware. As for CPU cores, you have to consider the peak CPU usage, not the average usage. Additionally, if you are using any of your VMs as backup proxies with Veeam (recommended) then they will need adequate CPU performance or your backups will suffer.

 

CPU load due to NICs can be minimised through the use of 10GbE as it is a different protocol and utilises TCP offload to reduce the CPU usage. But, this also is a reason why I recommend bumping up the CPU configuration for VM Hosts. This is my experience from 7 years of virtualisation.

 

While I agree with the sentiment, the size of the environment says different. Personally I'd stick with local storage in each host server, but I think a couple of grand could buy a suitable NAS that support iSCSI with good enough speed/reliability for that environment. SAN's are rediculously expensive and I very much doubt is something you should be looking at.

 

I agree, sticking with local storage in the server or DAS is recommended here. However, I do not think a couple of grand for a NAS would in any way bring acceptable results. You are going to experience high (noticeable to end user) read and write latencies, and the CPU/RAM of a $2,000 NAS is not in any way going to be able to handle the workload. First of all, you wouldn't even be able to barely buy 5-6 RAID quality (WD SE or RE) drives for that amount of money let along a decent NAS unit to hold the drives. Using consumer-grade drives will result in tears (slow performance, RAID failures due to drives dropping off, and a higher drive failure rate than expected).

 

All servers support 10GbE - it's called PCIe. Seriously, don't buy this now unless you need it. Buy it when you need it, when prices come down.

 

Actually, that's not the case. PCIe comes in different widths, with single lane PCIe providing 2Gbp/s duplex bandwidth. So, a 4 lane PCIe bus will cater for 10GbE, but anything less will not. I have two older servers in use with 4 PCIe slots that do not support 10GbE. Most new servers do support it, but some lower to mid-tier servers do not.

 

I suppose the point I'm trying to make is there is a danger of over spec'ing and spending too much (or getting a project mothballed on cost), as much as there's a danger in under spec'ing.

 

It is my experience that underspec'ing is done far more often for a virtualised environment than they are overspec'd. This happens much less frequently in a physical environment because it is a more familiar environment to most techs. The comments I'm seeing on here about the use of single server VM environments pretty much confirms that.

 

Other thing to bear in mind - Hyper-V pretty much requires atleast 1 physical domain controller (I'm aware 2012 R2 has "measures" to get around this particular chicken and egg, but...)

 

Yes, it's still recommended to have a physical DC (AD and DNS) or some things can break if your VM servers are all offline. However, a low-end Dell server will suffice for this. We use an R210 with 4GB RAM housed in a separate building from our primary server infrastructure. This is also where I store my secondary Veeam backup server (DIY FreeNAS server using Lian Li case small flash drive for OS and 4 x Seagate Constellation drives for data. It is also small enough to carry offsite in a hurry like your QNAP. We also have off-site backups and a fireproof drive backups are stored in as well.

Edited February 3, 2014 by seawolf