OK, so you’ve virtualised a bunch of servers and saved yourself a bit of money on hardware. Life is a little easier because you no longer have to go through the server procurement and provisioning cycle quite so often to meet new requirements. But has anything fundamentally changed in the way you manage systems and deliver IT …
But not in the way IT planned
As a technical end-user, a high-performance 64-bit VM gives me the ability to actually get work done under Fedora on my locked-down, heavily-administered Win7 machine. Allocate all four cores and most of the memory to the Linux VM, and the only remaining reason to deal with Windows (and IT's idiotic policies like forcing home directories to be on intermittently-available servers halfway around the world) is (ugh) Outlook.
I have a few SBS 2011s that I look after, but one that ran hyperV to host a virtual XP actually wiped out 2 days worth of incoming emails. Turns out that all the networking settings were changed in a cascading effect. very odd but after removing HV all is happy again.
Im not a huge fan of industrial strength VM systems, too much to go wrong on a host that wipes out all the guests
I can see the day when everything comes back to physical. VM has a use but to cram everything into one box is fine until it goes wrong. At last over a couple of servers you can keep running.
Re: VM iffy
Well too often virtualisation is done on the cheap.
In any business environment you should never had a single host with multiple important VM's on.
You need multiple high avaliablity hosts, high resiliant SAN's and networking. Basically if you arent spending at least the same ball park on hardware as you would if everything is physical you are doing it wrong.
The biggest advantage of virtualisation isn't saving on hardware costs, it high avaliability, reduced administration in building/patching servers , saving on power and space
Re: VM iffy
hyper-v is not supported in any way, shape, or form on sbs 20xx. someone was asking for trouble and ignoring a wealth of documentation when they implemented it.
Its another layer for the support team to take care to. Rather than software to hardware mapping, we now have software - hypervisor - hardware.
As mentioned above, virtualisation provides many benefits when it comes to high availability. However, while virtualisation provides all the benefits mentioned in the article and many more, it requires its own support to maintain the infrastructure in a state that will support the benefits its supposed to offer and quickly becomes a hinerance if things aren't up to scratch or fail.. Not to mention, a hardware failure could take out multiple services.
Just move everything to one of the commerically available cloud service providers (which of course has its own set of problems)!
It's not supposed to end all our worries
However it does have its advantages and saved me many hours of work on a server where I didn't need to reinstall a Windows system and was just able to copy an image over.
At work it allows me to use the legacy software we have there (from 1998! written for Windows) while still having a usable Ubuntu Linux system to actually get some work done.
The problem we have is persuading non technical decision makers of the merits and possible uses for the technology they've already paid for. We use HA/DRS VMware 4 clusters across multiple datacentres. Our SAN infrastructure is VMware aware but is not configured to use the functionality. We still need to 'ok' simple things like VMotioning a guest from one host to another during core hours due to the "risk". When you try to explain that we have DRS enabled on the cluster and therefore VMotion happens every now and then without us doing anything you just a blank look, followed by a "no, wait until a downtime period". In my experience, the technology is pretty sound. People aren't using it properly or don't trust it enough to make the most of what it has to offer.
It does what it should when done right
In my previous job, going to virtual (in the ESX 2.5 days), meant after the initial heavy lifting, we could cut the server hardware budget to 1/5... and it stayed there year after year. And as we also see in my new job, the virtualized platform provides a far more robust base for handling special conditions, from capacity changes to unplanned downtime. we can now take down a server for maintenance during work hours and everything just stays up, even all the (majority of) apps that can't be clustered using Windows clustering.
Everywhere I've been it's been a game changer, and I still sometimes catch myself smiling when I do things I would have considered magic when I started in this industry. Be it vMotion, automatic provisioning, or FT...
A Managerial View
The benefits of virtual over physical, from a hardware cost point of view, are real but are really only marginal. The real savings come when it is necessary to change the hardware whether due to h/w failure, performance issues or h/w support issues. As usual, the developers' interest in their application lasts only until application go-live date. After that, it's the Data Centre's problem! So try to arrange a h/w migration when the developers couldn't be arsed! With virtualization, h/w migration is a non-issue and can be done with little fuss and cost. So, for me, virtualization is a no-brainer and delivers high-availability and business continuity for free.
My 2 Cents
The advantages as I see it are: VMotion, hardware utilisation, multiple OS on one machine, ability to keep legacy systems running on modern hardware (the old "don't touch that server"), and convenience when replacing hardware. With regards needing better infrastructure (SANs etc) I think it is forcing you to do what you probably should be doing anyhow.