An Overview of Virtualization Technology

Jane Walker writes to tell us that TechTarget has a short writeup on virtualization and some of the ins and outs of using this technology effectively. From the article: "Virtualization is a hot topic in the enterprise space these days. It's being touted as the solution to every problem from server proliferation to CPU underutilization to application isolation. While the technology does indeed have many benefits, it's not without drawbacks."

Comments

Good reading until the end

[tinkertim]

From TFA:
>>>>>
  Novell is investing lots of effort in optimizing Xen specifically for running a virtualized copy of NetWare on top of Linux. The company's goal is to provide its customers with a migration path over to the Linux platform without giving up NetWare.
>>>>>

One of the many un-sung uses for Xen is a swiss army SAN. I'm glad to see someone touch on this.

>>>>>
If you want to use Linux as your host OS, you'll definitely have to go with VMware.
>>>>>

That wasn't so cool. I appreciate the fact that there are just too many products available to touch on everything in one short summary article / writeup, and while the majority of the article was informative even to the lay person, you need to end a sentence like that with a 'Because .... [summary]'. That's a really broad and sweeping statement to make.

Or perhaps even "I recommend VMWare" would have been better.

It looks like the author lost interest in what they were writing near the end of the article. They talk about IRC or newsgroups being the only support options available for OS products [another sweeping statement], however have you checked out the wiki at xensource.com lately?

Just seems like TFA lost coherency after 'What's best?' It went from really informative to misleading rather quickly. If your going to go to a virtualized platform you owe it to yourself to spend a month trying each candidate to see what works best for you, not the author of whatever article you read :) This is not a pro Xen rant but I'd like to point out that it does install effortlessly on most Debian systems in under an hour, the TFA sort of indicated otherwise.

Re:Good reading until the end

[Kangburra]

Just seems like TFA lost coherency after 'What's best?' It went from really informative to misleading rather quickly. If your going to go to a virtualized platform you owe it to yourself to spend a month trying each candidate to see what works best for you, not the author of whatever article you read :) This is not a pro Xen rant but I'd like to point out that it does install effortlessly on most Debian systems in under an hour, the TFA sort of indicated otherwise.

That's when the phone call from Redmond came through.

Re:Good reading until the end

[Anonymous Coward]

Well, if this is good reading, I can produce a lot of it:

"While the technology does indeed have many benefits, it's not without drawbacks"

unbelievable!!!!

Re:Good reading until the end

[tinkertim]

While TFA did annoy me at the end you have to admit its a very broad and difficult topic to cover in a way that pleases everyone. Most similar articles I've read get lost just in the 'how it works' section.

People who understand Virtualization aren't going to rely on this type of article to make decisions, they tend to prefer whitepapers and case studies. Someone without the benefit of experience is going to lean heavy on Google to find a solution.

There's just no way to cover everything that can be covered without losing the average reader. This one didn't go too terribly in depth and touched on some good points and really tried to educate the reader, up until the end.

So good reading, for this type of article.

Re:Good reading until the end

[Anonymous Coward]

I found the entire article to be almost content free. This is the sort of thing that PHBs like to read to feel they're "staying in touch" with technology.

For example, there was no mention of the reason behind the performance differences between VMware (ie. you're emulating everything, right down to the CPU) and Zones (ie. you're running one kernel and only jailing processes).

Fortunately, it was short enough that I could get to the end without wasting too many seconds of my life. But the /. keepers really ought to think twice before posting articles like this, even on a slow news day. It does nothing to improve the image of the site.

Re:Good reading until the end

[IO ERROR]

Looks like you took at least one of those quotes out of context. Here's the context:

If you're a developer looking for a flexible way to test your application in multiple environments, you'll probably want to go with either Virtual PC or VMware Workstation. If you want to use Linux as your host OS, you'll definitely have to go with VMware.

Virtual PC doesn't run on a Linux host, so you'll definitely have to go with VMware.

Re:Good reading until the end

[Stringfellow]

This guy is obviously a Novell shill just trying to get them some more visibility. I like this line: "Supportability is another major concern that could be set aside with a vendor like Novell standing behind an option like Xen." Yeah, I'm going to jump right on Xen 'cuz Novell is behind it? How many times has Novell gotten behind something long enough to sell its wares and then dumped it for the next greatest thing?

Re:Good reading until the end

[tinkertim]

I re-read it a few times just to be sure. Its not very well constructed, but I think I got the context right in my post .. I think. Read it this way swapping a space with a new line:

If you're a developer looking for a flexible way to test your application in multiple environments, you'll probably want to go with either Virtual PC or VMware Workstation.

If you want to use Linux as your host OS, you'll definitely have to go with VMware.

The way that reads to me is, if you want to use Linux as your host (root) OS , you have no choice but to go with VMWare. What about Xen?

You almost need to read it out loud. Not the best constructed article. I had to do a couple double takes just to be sure after reading your comment .. but I think I got the right meaning and didn't twist the context.. wasn't my intention to do so. It seemed pretty clear the first time I read it, unless I'm missing something? I am rather dense prior to my 6'th cup of coffee and only on #4 right now so please do reply if I'm way off base ..

Re:Good reading until the end

[Torne]

You can't run Xen on Linux - Xen is a freestanding hypervisor that runs directly on the metal. So, the statement in the article seems perfectly reasonable. (this is, incidentally, one of the (numerous) advantages of Xen over VMWare in the performance stakes - being able to control the hardware directly instead of having to mess around with what the host OS will let you do is quicker).

Every OS on a Xen system is a guest OS. Some of them just have permission to create new OS instances, or access particular bits of real hardware directly.

Re:Good reading until the end

[Hercynium]

Xen is a freestanding hypervisor that runs directly on the metal.

I feel I should get a little pedantic with this statement. Xen, specifically, is a modification of the linux kernel that provides hypervisor capabilities for the host OS (Linux) and integration with the guest OSes. Xen's host components can't run directly on the metal, like VMWare ESX can. It needs the rest of the linux (I can't remember if it's been ported into other kernels) kernel to provide hardware access. Also, Xen requires (until pacifica, et al) that the guest OS kernels be modified to integrate with the host's hypervisor layer. Without that, Xen does nothing. (VMWare ESX does not require a modified guest OS)

So, if you think of a Xen-enabled linux kernel as Xen, you're right. But I see it as a seperately developed, ported, and integrated extension that requires a kernel to operate. Again - I believe there are efforts to get it running inside other kernels, but I don't remember.

On second thought - someone fill me in here - I'm guessing that VMWare ESX probably runs as part of some ther's OSes HAL, but of course they don't say so in the sales pitch...

*disclaimer* -- While I really like VMWare's product for functionality and ease-of-use, for performance I'd go with Xen. I'm currently involved with a project at my company to virtualize as much of our datacenter as possible, and I've been pitching Xen to the group, over VMWare, provided XenSource's product lives up to it's marketing specs.

Re:Good reading until the end

[tinkertim]

Correct. I work with Xen daily and most of my products and services are built around it. One of which is a replacement for Virtuozzo for the purposes of maximizing and isolating resources or web hosting companies.

Xen augments the kernel, it does not replace it. The Xen hypervisor then interacts with the host (dom-0) kernel.

dom-u (guest) images can then boot using any kernel modified to interact with the Xen hypervisor. Currently we play with:

Debian (Sarge)
FC4
CentOS 4
NetBSD

As dom-u's (guest) OS's.

We have also enjoyed some success but not 100% stability bringing Win2k3 up as a dom-u.

I have deployed clusters that use Xen as a management layer and I can tell you, it *does* live up to its marketing specs. Xen's bridging is the fastest most efficient layer available, bar none. Its also a wonderful tool in helping to integrate a centralized storage area network into any size network and let people keep all of the protocols they like.

A *very* good source of information about Xen, what it does, how it does it is available on the option-c wiki (Here) , they also have some ready to go Debian installers that make installation quite easy (apt-get able).

Xen + OpenSSI is another fantastic combination if you take the time to really understand the networking possiblities and set it up appropriately. Good luck with the bean counters .. the price is right :)

Re:Good reading until the end

[yuriismaster]

Close... but here's the difference between Xen and VMWare ESX:

Xen does 'para-virtualization', wherein it virtualizes MOST of the hardware, but allows some passthrough to the bare metal. This requires virtualization-aware kernels and modifications to some software, perhaps. Since it's a 'lighter' application than the ESX server, it should run a bit faster.

VMWare does a full virtualization of every hardware component, like most other virtualization products (Virtual Server 2k3, Virtual PC, VMWare Workstation (I'm sure there are others, and I really don't know the field that much), so there is no requirement for the guest OS's (besides x86 and support for the 'VMWare Hardware', which is rather standard). VMWare provides the VMWare Tools package to Windows and Linux servers to enable better communication with the 'hypervisor' kernel, increasing network utilization, etc. However, it is very possible to run other OS's using the standard drivers.

Basically, if you're really going for the speed and don't mind the extra leg-work to get a working kernel in Xen, it should be faster. Since we have to run Windows 2003 servers, and really don't have the staff to hack around the kernels (not to mention already bought support for ESX), ESX runs beautifully for us.

Re:Good reading until the end

[yuriismaster]

And in my previous comment, I forgot to answer the 'second thought'.

ESX server is a modified Redhat kernel (2.4.9) and the ESX-specific apps, plus device modules to connect to whatever server-class hardware you have.

Re:Good reading until the end

[Anonymous Coward]

"dom-u (guest) images can then boot using any kernel modified to interact with the Xen hypervisor."

Yes this is how it works if you don't have hardware VT or Pacifica support.

If you do, unmodified kernels for any x86 OS can be run. (including windows)

Re:Good reading until the end

[Torne]

Actually I'm the one being a pedant. It's not a modification of the Linux kernel at all, though older versions did borrow heavily from the Linux device driver code - the Xen hypervisor is a totally seperate chunk of code and would theoretically be capable of running without a Linux in sight at all, as long as some other guest was able to perform the tasks of domain 0.

Xen's host components *do* run directly on the metal, they just don't talk to I/O devices. Xen controls the processor and memory, and provides inter-domain communications. It is an OS in its own right, just a very, very limited one that can do nothing other than schedule and communicate between its guests.

The original version of Xen included the device drivers for the real hardware in the Xen hypervisor, making domain 0 even less special - domain 0 accessed the hardware through the virtual device layer the same as all the others. This was changed to the current model to avoid having to maintain a complete device driver layer in the hypervisor.

Domain 0 is *not special*. It has been given permission to start other domains (but you can give that permission to any domain, not just dom0) and it is by default given permission to talk to all the hardware (but you can give selective rights to access particular bits of hardware to any domain - you could even run a seperate domain for each device driver, so that a bug in the SCSI driver can't affect the network driver, for example).

You totally misunderstand the Xen architecture ;)

BTW, I worked for the Xen development team at Cambridge University, working for Ian Pratt, who is now the Xen project leader at XenSource. So I really should be familiar with its design...

Re:Good reading until the end

[Hercynium]

I appreciate the correction, and the credentials. Xen is an exciting technology to me so I have a tendency to jump in to the discussion. My only defense is that there is *a lot* of inaccurate information available, even from what would normally be considered reliable sources.

I hope my employer will choose Xen as their virtualization platform, but I know that much of that decision will hinge on whether or not Windows can run stably. Since the decision won't be made until next quarter, I'm crossing my fingers for a supported solution to that quandry. (My company tends to prefer 'traditional' commercial software, but they're not afraid of open source type stuff, as long as there's going to be long-term support - I'm pushing XenSource)

Re:Good reading until the end

[Torne]

I was working on porting Windows XP to Xen (without the benefit of VT/Pacifica, which is now the 'right' way to do this - it was some time ago, before those techs were announced). It was kinda fun ;)

Got to stick my hands in the Windows source, and futz with it. It got quite far but it was never what you might call 'stable'. It also took about 15-20 minutes to do the first few seconds' worth of booting due to the vast, vast amount of debug information we had it dumping out of the serial port. 295MB bootlogs anyone?

With VT/Pacifica it should be a breeze. I'm not actively involved in developing Xen any more (though I do use several Xen machines), but it's a nifty project and works impressively well (even if people don't actually understand how the hell it works *grin*)

Psst. btw

[BadAnalogyGuy]

Microsoft has made their server virtualization software available for free.

http://www.microsoft.com/windowsserversystem/virtu alserver/software/default.mspx

Maybe... but it's no XML!

[Siffy]

or Java!

Shameless plug

[digitalhermit]

In South Florida tomorrow (Thursday), a dorky looking guy will be presenting an introduction to Xen talk. Check http://www.flux.org/ for details.

Windows Licence Issues. (wrt. Virtulization)

[Domini]

One thing the article does not speak about is licensing issues when using Virtulization. For instance MS has some twists and turns...

For instance:

One needs 2 different licenses if you run XP in XP.
You can run 4 instances of Windows Server for free in Windows Virtual Server.
You can run one copy of an older windows for free in Windows Vista.
(You can read more about this on the MS site...)

For Windows XP General Purpose license User Rights:

http://www.microsoftvolumelicensing.com/userights/ PUR.aspx

Download and read document, section "Microsoft Desktop Operating Systems" which reads:

  I) Installation and Use Rights.
    a) You may install up to two copies of the software on one device.
    b) Except as provided in Section II.a and II.b below, only one user may use the software at a time.
    c) You may run a prior version in place of the licensed version for either or both of the copies.
    d) You may only use the copies on the device on which you first install them.
    e) You may use the software on up to two processors on that device at one time.

Thus this means that I can install and use XP as Bootcamp native and Parallels VM guest using only one license.

yay!

Re:Windows Licence Issues. (wrt. Virtulization)

[Anonymous Coward]

You do understand that unless you sign a contract with Microsoft or otherwise make a binding agreement, you have your full "fair use" rights to do with purchased software what you please. Thus, you are not restricted by Microsoft's verbiage found on its website. Don't be fooled by EULA's. They are not enforceable (i.e., binding) unless you agree to them before you make the purchase of the software.

Put differently, if you are the only one using a book you purchased, you can make as many copies of that book you want for your own personal use. And you can attempt to read them all at the same time, if that floats your boat. The publisher has no say in it, unless you signed a specific contract beforehand limiting your usage.

Re:Windows Licence Issues. (wrt. Virtulization)

[Anonymous Coward]

those licensing doesnt apply only to Microsoft Virtual Server,
check this microsoft paper: "Licensing Microsoft Server Products with Microsoft Virtual Server R2 December 2005. and Other Virtual Machine Technologies"
http://download.microsoft.com/download/6/8/9/68964 284-864d-4a6d-aed9-f2c1f8f23e14/virtualization_whi tepaper.doc

Re:Windows Licence Issues. (wrt. Virtulization)

[Decker-Mage]

Frankly that's a can of worms that I stopped worrying about years ago as I have more licenses than I know what to do with. However, I would never run XP as my base system. Too much consumer crap in the way tying up system resources. Always select a server OS, even one of your older Windows 2000 Server or Advanced Server, as your base OS and disable as many services as you don't need, if you use Windows as your base OS. You'll see far better performance with either VMWare or Windows Virtual Server. I know, I've done the benchmarks.

Re:Windows Licence Issues. (wrt. Virtulization)

[dbc001]

> a) You may install up to two copies of the software on one device.

I'm a web developer and would like to start testing my sites with Internet Explorer 7 (which is currently in beta but overwrites IE6). Unfortunately MS requires Genuine Windows Advantage validation to download IE7. Can I validate both my host OS and my virtual guest OS with the same CD Key? Has anyone else encountered this problem? Am I the only one who can't stand this CD key validation crap?

(just kidding about that last question, i know the answer. the rest was serious though.)

Re:Windows Licence Issues. (wrt. Virtulization)

[Abcd1234]

Well, you can say that all you want, but until a real life EULA is invalidated in a court of law, I wouldn't be taking my chances, unless you feel like being a legal guinea pig.

Re:Windows Licence Issues. (wrt. Virtulization)

[Anonymous Coward]

Eh, centuries of common law contract law say otherwise. Agreements are real things. EULA's, in spite of the acronym expansion, are not agreements. They are written parts of the copyrighted work you have already purchased the right to read. Reading the preface of a book you own and then continuing on to the first chapter, even if that first chapter is behind a lock, does not make you enter into a contract. You already have the right to read that first chapter. You bought the book.

There is a reason this does not get meaningfully tested in court. And it is not because the companies with billions of dollars in liquidated cash would win.

Re:Windows Licence Issues. (wrt. Virtulization)

[Domini]

From what I could understand from reading those licence agreement files, you should be allowed to install it twice... whether you are allowed to run both at the same time may be a different matter.

I think I read someplace that the same licence key may not be present on a network on two hosts, but I'm not sure in what context I read that...

Perhaps download the doc I linked to earlier and read up further?

And good luck with that IE7 beast!

Re:Psst. btw

[Anonymous Coward]

Yeah, well my dad can beat up your dad. Plus I heard Microsoft's "virtualization" has no hypervisor, so isn't really virtualization. And VMware gives theirs away for free, too. So, what were we talking about? Our sisters?

Re:Psst. btw

[Anonymous Coward]

You haven't got the slightest idea what you're talking about.

Re:previous hot topic: virtual reality

[Ohreally_factor]

It's for this reason that virtualization is mostly hype, and won't be accepted in the enterprise. IT departments have better things to do with their budgets than to buy fancy VR goggles and data gloves for their admins.

Piss Poor Article

[HoofArted]

Not much in it, is there? The outcome ... "it depends".

Re:Psst. btw

[Decker-Mage]

I agree with AC and I will post by name. You haven't got a clue.

Re:Psst. btw

[Decker-Mage]

With VMWare Server (ex-GSX) switching to free status, frankly I don't think they had a choice. I've been working with, and beta-testing for years, with both and the VMWare product still wins in my opinion. No win situation for MS.

Re:Psst. btw

[LarsWestergren]

Psst yourself.

You have to pay for the OS to run the virtualisation server on, you have to register to download it, and then you have to follow the usual licences- i.e (From MS own Virtual Server 2005 Technical Overview White Paper):
* you may not transfer original OEM server licenses from one computer to another,
* Each installed copy of Windows Server must be separately licensed. This means, for example, that if you are setting up four virtual machines within Virtual Server 2005 to run one instance of Windows 2000 Server and three instances of Windows NT Server 4.0 concurrently, you will need one Windows 2000 Server license and three Windows NT Server 4.0 licenses, in addition to the Windows Server 2003 host license running Virtual Server 2005.
* Each additional licence such as for IIS or databases have to be paid for each virtual machine... and so on.

Yeah, that sounds like an awesome deal.

Strange

[DrMrLordX]

The article doesn't even touch on Intel's VT or AMD's Pacifica technologies. What gives?

Re:Strange

[Malor]

Yeah, I was thinking the same thing. This article would have been interesting, say, 18 months ago... but with VT and Pacifica, things are different now. Without at least mentioning those, it's not very useful.

Anyone have a pointer to a good writeup on the differences between VT, Pacifica, and regular old software virtualization?

Re:Strange

[tinkertim]

>>
  The article doesn't even touch on Intel's VT or AMD's Pacifica technologies. What gives?
>>>

Or being able to have win2k3 happily running as a dom-u under Xen 3 on such hardware. There's only so much you can fit in a one page blurb however. And the average reader wants it all compressed into 5 minutes or less of reading.

Topics like this , you just can't do that unless you link to many external resources allowing the reader to get more about whatever interests them.. which is what TFA should have done.

But it comes quite a bit closer to being a down to eart overview than many others on the same topic that I've read.

Re:Strange

[rikkus-x]

What I'm most interested in:

• Where and when we can buy chips with these features - no-one seems to know for sure if the new Intel Macs have VT enabled.
• What sort of speed boost they give - if it's not too huge, it's not such a problem to buy chips which don't have the features enabled.

Rik

Re:Strange

[Anonymous Coward]

no-one seems to know for sure if the new Intel Macs have VT enabled.

They seem to, other than (apparently) some Mac minis. However, some people have had good luck in reenabling it even on those. See this thread:

http://forum.parallels.com/thread85.html

Re:previous hot topic: virtual reality

[Decker-Mage]

If you play in the Enterprise space, virtuationization can be a god-send, just on the basis of server consolidation alone. Where I play, systems security from SMB to Enterprise, I've been doing some interesting work on developing bastions (no relation to the Linux setup with the same name, think the Military Engineers Vauban and Michelangelo) that I think will play well once fully worked out. Heck, even on a consumer machine, if all internet work is done in a VM, you cut the risk of infection by a wide margin. Those are just a three sample ways to go with this technology. Virtuationzation just gives you more interesting capabilities.

Just my $.02

Macwintelintosh

[infinite jester]

Since Apple's move to Intel processors and the recent releases of Boot Camp and Parallels Workstation, running Windows within or beside Mac OS X is suddenly all the rage. My question is, has anyone thought to use Boot Camp to load Windows, and then use a Windows virtualization solution to run OS X inside of Windows-on-a-Mac? I'd try it myself, but I fear that I might rip a hole in the fabric of space-time.

Re:Macwintelintosh

[Anonymous Coward]

I was just thinking to myself what a nice change a slashdot article that isn't about bootcamp but then....

Come on its only a boot loader what is the big deal

Re:Macwintelintosh

[Anonymous Coward]

OSX only installs on Apple (Intel) hardware, right? So I guess it wouldn't work, as VMs present their own virtual hardware to the guest operating system.

Re:Macwintelintosh

[kscguru]

OSX inside a VM?

Alas, illegal (at least in the US). Mac OS X has some trusted computing code that verifies that it is running on Apple hardware. Bypassing this code is a definite DMCA violation (hence no company will try it). AND Apple assumes a very narrow set of underlying hardware, which isn't what VMs provide - so there is another pile of effort to emulate chipset and so on.

In short, it's not going to happen until Apple wants it to happen.

Re:Psst. btw

[BadAnalogyGuy]

While you are correct about the cost of the physical server software, I don't find anywhere in the licensing agreement that says I can't run some other OS. In fact, I am running 4 VMs with various Linux distros installed as we speak.

And as for those licensing restrictions, they will apply to any VM software that runs your Windows Server OS. So what's the solution? Well, don't use it to run Windows, for one.

Re:Psst. btw

[Anonymous Coward]

Your criticism is straight forward and cogent. I don't know how I can offer a counter-argument. No, really.

Netscaping

[Savage-Rabbit]

Microsoft has made their server virtualization software available for free.

Isn't this the opening phase of what Computer Business Review calls 'Netscaping' the competition? I wonder if that word will ever make it's way into the Microsoft system spelling dictionaries?

Re:Netscaping

[BadAnalogyGuy]

Well, to really be Netscaping, VMWare would have to put out a completely unusable version 4.

Re:Netscaping

[Anonymous Coward]

Well, to really be Netscaping, VMWare would have to put out a completely unusable version 4.

I was there when the original 'Netscaping' unfolded. Netscape Navigator 4.x sucked but it wasn't a totally useless browser either. To be fair to MS what killed Netscape wasn't just the free competing Microsoft product, Microsoft deserves some credit for the fact that with time IE simply improved to the point that it sucked less than Netscape. Fortunately we now have a variety of browsers to choose from that are better than IE.

Re:Netscaping

[Anonymous Coward]

It was a better browser than Netscape. The other fact that aided it, was it was installed by default. There was no need to search it out and download it.

The virtulization server market is a LOT different than the web browser market. People who need a virtual machine know enough to look around at options and not just use what's pre-packaged with an OS. I don't think there are worries about MS killing the VMWare/Xen market because of ubiquity.

Re:previous hot topic: virtual reality

[BadAnalogyGuy]

If you play in the Enterprise space, virtuationization can be a god-send, just on the basis of server consolidation alone.

Yeah, and also because you can use the virtualization software to freeze the CPU which allows you to finally defeat those Romulans in melee combat.

Application Virtualization

[cheeseflandan]

FTA: "If you're trying to solve one of the server-based issues like consolidation or application isolation, you'll want to go with a server solution"

Hmm - I think there are a few vendors who'd disagree with that.. Softricity, Altiris, Citrix, Wise to name a few..

Re:Application Virtualization

[scapaman]

I don't know. I work in this field, and see people dropping those technologies and moving to virtualisation. Now that hardware is powerful enough to support it, why not just roll out 32 desktop XP images with all apps installed, rather than dicking about trying to get an app installed on Citrix or Softricity?

Re:Application Virtualization

[cheeseflandan]

That's a good approach, and I definitely see the value in it (and recommend it in some cases), but then you have the questions of how to get the apps onto those virtualised machines? Yes, you can use some form of ESD, or even manual installations, but then what about when you have apps which 'conflict', or require constant updates? And what about users who require off-line access to their apps, etc?

The point I was trying to make was that the article completely ignored the areas of app "virtualisation", not necessarily that it's the right approach for everyone..

Re:Application Virtualization

[scapaman]

very true. the sites where I see it use slipstreamed images to keep updated. Offline access is not usually an issue (for offshore call centre staff!).

Re:Application Virtualization

[utlemming]

With VMware's ACE product you can exaclty that for an organization and lock 'em down.

I run Linux as my host with VMware. Frankly, I love it. I have probably a dozen Windows images that I use each with a slatted purpouse. I even run Oracle (and it runs quite well at that) in it's own VM. The best part about doing this approach is that I am able to isolate programs that I really don't want running together, like keeping Oracle seperate from my host. Also, I have a clean enviorment where I can play with malicious programs or run programs that I think might be suspect. I can honestly say that I have never enjoyed that stability of Windows like I do right now in a VM.

Re:Application Virtualization

[Anonymous Coward]

The Altiris product (Software Virtualization Solution) is managable via any framework. You can use command line strings, WMI calls, or API Calls to manage the installation and activation of the virtual packages you create. You could even use login scripts if you don't own an Enterprise Software Delivery tool. If you do, Altiris has GUI's that snap into the Altiris Console, and there is also an MMC snap in Microsoft's SMS too. However, it's not required if you choose to code your own automation or use command line strings that you somehow execute remotely. There's even a user based self service tray app that allows you to reset and activate and decactive layers too, if you trust your users with that much functionality once the virtualized apps are installed.

Conflicts aren't an issue because each install is sandboxed. You can litterally have multiple versions of the same file in the same directory. Each application uses the version it installed, and doesn't touch the version another app installed. It's similar to a traditional type of application isolation (DLL redirection) but isn't limited to DLL's. The entire application is isolated to a hidden redirection directory, including all it's registry keys. The app executes normally, except that the virtualization driver redirects file requests to the hidden areas of the machine.

Offline access isn't a problem. Altiris SVS installs to a local hidden directory, there is no server required or any other communication over the wire to make the virtualized app work. Citrix and Softricity use radically different approaches. SVS installed apps behave normally, and in most cases the user never knows they are using a virtualized application.

I agree, it is dissappointing that the article doesn't mention this new method of virtualization, which works on older hardware, is extensible via any framework, and does not change the end user experience.

Oh, and it's free for personal use:

http://www.download.com/Software-Virtualization-So lution/3000-2651_4-10516818.html

Re:previous hot topic: virtual reality

[Decker-Mage]

LOL! Weeelllll, you do have to be careful about your mix of applications that are running in each VM OS. Seriously, if you are that worried about system freeze, VMWare ESX is the way to go and you need to step-up to some serious hardware from, say, IBM or one of the other big players. Then again, with a four-way, dual-core Opteron setup, you can get some pretty awesome VM set-ups and spend a lot less money than you would have a year or a year and a half ago. That's what I'm evaluating here for my VM server configuration, either a two-way or four-way, dual-core Opteron rig. Frankly, the prices have come down rather extremely of late. [And to think I used to be such a big Intel fan-boy for a decade. Pfffft, Intel.]

Re:Psst. btw

[Decker-Mage]

Your MS licensing information is out of date. They've changed the way they handle Server 2003. Furthermore, you don't have to use Server 2003 as your base OS for VS 2005. I was using XP Pro SP2, Win'2K Server and Advanced Server, as well as Server 2003 Enterprise during the betas for both VS 2005 and VS 2005 R2. All worked just fine. Actually, I got the best performance from Win'2K AS after I really locked down the services running although that may be somewhat biased as I really know AS best and I didn't lockdown Server 2003 Enterprise. The improved memory model for Enterprise just might give it the edge if it were similarly configured.

Re:Psst. btw

[dc29A]

I virtualized a Windows NT4 IIS server running an ASP application with some VB COM components, VMWare ESX is incapable of running it without insane CPU usage. A one CPU physical server is running at about 30 tx/sec with 15% CPU usage, a virtual server inside ESX is getting 90% CPU usage with barley 5tx/sec, the VMWare host itself is at 65% CPU usage with 4 CPUs.

VMWare seems unable to deal with many object creations and many context switches, the application basically creates a COM object, deals with it and deletes it, very simple logic. A bit disappointing that VMWare is taking such a huge hit.

what a horrible article

[jnf]

Seriously, how did this make it on /.? The article is only a few paragraphs long, doesn't really even touch on hardware virtualization support or why its necessary (because virtualization currently sucks under 'normal' intel architecture). It even refers to qemu as virtualization, which its not, its an emulator. It mentions the program once then never touches on it again. It never explains why a person might want to use bochs or qemu even though its much slower than vmware/virtual pc. it doesn't touch on parallels or any other software out there.

Even more it doesn't even explain why the suggestions it makes are made. This article is basically a badly written advertisement for vmware or virtual pc.

Re:what a horrible article

[ettlz]

It even refers to qemu as virtualization, which its not, its an emulator.

Technically correct, but maybe TFA was thinking about kqemu as well (which does provide something akin to virtualisation).

Re:what a horrible article

[jnf]

I fail to see how it was 'thinking of kqemu, furthermore I fail to see how a front end to an emulator makes it akin to virtualization (thats all kqemu is, a front end)

Re:what a horrible article

[ettlz]

I thought kqemu, the kernel module, allowed user-space processes to run natively on the host CPU, wereas any supervisor-mode instructions are run under emulation.

Re:what a horrible article

[jnf]

oh crazy ;] kqemu is also the name of a kde front-end ;] sorry, i stand corrected ;]

Re:what a horrible article

[runderwo]

Hooray for overloaded open source project names!

Informative but you left out one important thing..

[bigtallmofo]

You can run 4 instances of Windows Server for free in Windows Virtual Server.

This is only true if you are running Windows Server 2003 Enterprise edition. And the host operating system counts as an instance. So, you load Windows Server 2003 to act as a Virtual Server host, that means the license allows you to run 3 additional guest operating systems within that host.

These are two important distinctions because the Enterprise edition is far more expensive than the standard edition. It wouldn't make sense to purchase it just for the 3 additional licenses.

The article is an outdated joke.

[bigtallmofo]

I agree with you and even take it a step further. The article could not be more plain in that even though it was dated 4/7/2006, it did not take events of the last month into account which makes it totally useless in my opinion.

Three major announcements in the last month have radically changed server virtualization and made the article obsolete:

1. VMWare renamed GSX to Virtual Server and made it free.
2. Microsoft made their Virtual Server free.
3. Microsoft announced support for certain Linux distributions in their Virtual Server product.

The parts of the article that show it's obsolete in light of the above facts:

An open source solution will win the cost battle almost every time
If you want to use Linux as your host OS, you'll definitely have to go with VMware.

Also, for my own personal review - I'm a pretty heavy Microsoft user and was excited about them making Virtual Server free. Evaluating VMWare's free product against Microsoft makes Microsoft look pretty unpolished though. For instance, compare VMWare's P2V application to convert Physical to Virtual servers against Microsoft's offering which requires having a spare server lying around which must run Windows Server 2003 Enterprise with Automated Deployment Services. Give me a break - the cost becomes so prohibitive it's not even worth it. Microsoft may get there but right now their product looks like what it is - a bunch of things hastily thrown together. VMWare's products have coherence.

I think we've heard this before

[oddRaisin]

These claims (server proliferation, cpu/resource under utilisation) have been made before with utility computing. A company called Ejascent (later purchased by Veritas) offered utility computing software that very closely resembled Solaris 10's containers. And now, of course, Solaris 10 is offering containers natively. So the technology to consolidate servers has already been around for quite a while.

I realise that it's not quite the same -- virtualisation offers multiple operating systems on a single system while containers only offer multiple instances of a single OS. Virtualisation seems quite attrative. But I didn't ever see utility computing take off, or even the technology used to consolidate servers.

Virtualisation is a fun toy and may be a useful tool if you're a multi-platform developer. But it does not seem to be a serious enterprise solution for the datacenter. And talk about a patching nightmare! A virtualisation solution running on Windows with, say 5 instances of Windows. That's 6 copies of patches to apply, resulting in at least 11 reboots (1 for each instance and 1+5 for the primary OS).

The cons, administrative overhead and the resource overhead of running multiple images don't really add up to a significant cost savings.

Re:I think we've heard this before

[ratboy666]

Consolidation technology IS important. And it is "taking off".

Servers are more powerful now. If a company decides to consolidate physical resources (to save A/C, power, rack space, buildings), they can certainly "vertically stack" applications that used to run on multiple servers onto a single server.

However, if this is done with old-hat technology, the system becomes very difficult to manage. For example, I just worked on a 4 way Opteron with 8GB of memory. The NORMAL process list was 1800 lines long!

So, containers are used to segregate the machine into more managable units.

The uptake for this may seem slow, because the clients interested in this have to replace existing gear and facilities. We are talking about major facilities: one client has 7000 assorted Unix, spread across 6 datacenters to be consolidated into 1000 servers at 2 datacenters; another has 10,000 Solaris servers. It takes years to migrate these installations.

Ratboy.

Re:I think we've heard this before

[swb]

OK, let's say you have 5 Win2k03 applications that require 5 servers.

Your capital outlay for the servers will be around $10k. Add in a support package for hardware and you're talking another $500 or so per year. Each box uses power, requires a KVM interface, physical space, a network port and puts a heat load onto the air conditioner. These marginal costs will add another $100 per box, so we're now looking at a 3 year cost of around $13k, not including any labor costs.

A GSX rollout would have been about $7500 for a bigger server and a GSX license. Setup would have been faster, since the base Wink03 image could have been cloned 4 times. The single server would require less A/C, power, switch ports and KVM ports, but we can still call it $200 and be safe. Our three year cost would still be under the hardware only costs of the 5 server configuration.

Patching is really a non-factor, since you're patching regardless, and you can patch 5 VMs and then the server and really get away with one reboot, since most Wink03 patches don't require an immediate reboot.

Of course this barely takes into the ease of managing a virtual box (cloning machines for patch testing, backup of raw images, etc), the flexibility of being able to create a new server almost instantly without having to hunt down hardware. And once you get into the "datacenter" rollouts with SANs and multiple ESX hosts and the ability to move VMs between hardware platforms, it gets even better.

Re:I think we've heard this before

[VdG]

Agreed. We make extensive use of virtualization/partitioning technologies with IBM pSeries hardware, running various versions of AIX. We've also got some interest in virtualization for our Windows servers, but nothing's actually be done in that area so far: something for the future, perhaps.

For us, although there is an additional administrative load that's pretty minor and is vastly outweighed by the increased flexibility: being able to create new virtual servers or change existing ones at very short notice. It depends a bit on exactly what deals you can get, but for us the hardware works out cheaper, too: a couple of dozen really big servers are cheaper than a couple of hundred smaller ones.

Re:I think we've heard this before

[Anonymous Coward]

A virtualisation solution running on Windows with, say 5 instances of Windows. That's 6 copies of patches to apply, resulting in at least 11 reboots (1 for each instance and 1+5 for the primary OS).

Huh? Why would the primary OS need to be rebooted more than once? Are you saying that you can't reboot virtual instances without rebooting the primary OS? I find that hard to believe. If true, virtualization would be worthless, patches or no patches.

Re:I think we've heard this before

[Wolfrider]

> And talk about a patching nightmare! A virtualisation solution running on Windows with, say 5 instances of Windows. That's 6 copies of patches to apply, resulting in at least 11 reboots...

--This is why serious virtualization servers don't get run on Windoze. It's fine if you need to run Windows GUESTS - use a caching proxy server like Squid to download the patches, and stagger the automatic-update times for when they automatically reboot.

Re:I think we've heard this before

[LurkerXXX]

A virtualisation solution running on Windows with, say 5 instances of Windows. That's 6 copies of patches to apply, resulting in at least 11 reboots (1 for each instance and 1+5 for the primary OS).

6 copies of patches to apply? Um no. Any admin working with that kind of setup SHOULD know about WSUS server and be rolling out patches (after he's evaluated them on a test rig to make sure they don't break any of his company-specific software) automatically.

And no, it's not 11 reboots. That's a really really dumb way to do it. You set a group policy to prevent the machines from automatically rebooting after patch installation. When it's time for the scheduled maintenence you shut down all the VM's, reboot the host OS, then crank back up the VMs. That's a total of 6 reboots for 6 windows machines.

Virtualisation is a fun toy and may be a useful tool if you're a multi-platform developer. But it does not seem to be a serious enterprise solution for the datacenter.

Virtualization IS a serius enteprise solution. Lots and lots of us have it in production. Then again, we know a bit about the field and don't patch every machine by hand and do unneccessary reboots.

The cost savings are real if you hire someone competent to run the machines.

Re:Psst. btw

[Decker-Mage]

Agreed, there are situations that are totally unsuitable for virtualization. For instance, don't even think about a DirectX game under a VM instance which also exhibit similar behaviors in many ways. I do try it every time a new beta hits and am frequently disappointed.

Re:Psst. btw

[smithcl8]

You always need to pay your licensing fee for each Windows system running in a VM. The only difference between MS and VMWare licensing costs would be the cost of the Windows-based host system, which you would pay for if you were hosting your VMs on a Windows machine either way. Microsoft is getting their money no matter what. If you don't like it, don't use it.

Uses of virtualization for servers

[mpcooke3]

I have been using a few Xen based virtual servers from a commercial company recently - I used to manage physical machines. Here are some of my thoughts:

Advantages:
* Low performance overhead of Xen compared to other virtual solutions, and full OS level access as if it was a normal server.
* The cost of a hosted Xen solution is very low given that the hardware is usually managed.
* Reduced/No trips to the data center to replace hard disks etc,
* From the provider i use you can also reinstall the OS, snapshot and restore snapshots over a web interface and get access to the console. These are features you can set up in your own data center but most people never get round to.
* Quicker turn around if you need new servers, since normally they already have the spare hardware it's 1 or 2 days to get a new server set up rather than 1 to 2 weeks to order, install and configure it.
* You could do loadbalancing over several Xen Virtual hosts on physically separate machines very cost effectively. This would also mitigate against the variable performance on different Xen hosts if you used a dynamic weighting loadbalancer.

Disadvantages:
* Sometimes other users on the Xen system cause problems, or the server is restarted due to Xen related problems. This hasn't happened that often but you wouldn't currently run a system that needed 99.999% availability on a XEN virtual host if the system is vulnerable to a single server going down.
* You never know quite what your worst case performance is going to be like.
* If your system doesn't scale laterally (more servers) but only by buying a more powerful single server (some databases for example) then the Xen virtual hosting is not cost efficient.

Re:Uses of virtualization for servers

[ovz_kir]

You might try evaluating OpenVZ then -- to me it looks like it does not have any of the disadvantages you mentioned. If you plan it right and do not oversell, you can guarantee the certain quality of service to every virtual environment (VE). In addition, OpenVZ has dynamic resource management (you can't give more memory to Xen guest during runtime, can you? but it is trivial with OpenVZ) and much higher scalability (hundreds of VEs vs. tens of guests). As for the scalability across multiple servers -- I think nobody really do it in Intel PC world -- it's a real challenge to implement. But I think it will be done in the next 10 years or so. For now, OpenVZ can offer you a live migration feature -- when you need more resources than this very physical server can provide, you just migrate your VE to a better one, and then upscale its resource limits.

Re:Uses of virtualization for servers

[ovz_kir]

Quicker turn around if you need new servers, since normally they already have the spare hardware it's 1 or 2 days to get a new server

*shameless plug* if you own a physical server running OpenVZ, you can have a new VE (virtual environment, virtual server) in a minute. I'm not kidding -- it's just a few commands, and you are all set. Basically, you just have to choose which Linux distro do you want, and supply an IP address, name server, and the root password -- and you'll have your new shiny server in a minute.

Re:Uses of virtualization for servers

[mpcooke3]

I'm not hosting Xen i'm using a hosted solution that offers Xen based virtual servers. It only takes a couple of minutes to set up a xen server too if you are running the Xen host yourself - you can play with it yourself in the Xen live CD. The time lag i was referring to was dealing with a hosting company.

CoLinux

[radarsat1]

I noticed that whenever virtualization comes up, no one ever mentions CoLinux. I've tried it once and was quite impressed. It takes a different approach entirely--rather than running in a virtualized environment, it is actually a port of the Linux kernel to run as a Windows process. (Some hardware is virtualized by this method, however, such as the network interface.) Are there any advantages to this approach? In terms of reliability, speed, etc.?

Just curious.

Re:CoLinux

[Anonymous Coward]

Short answer: No.
Long answer: Windows sucks. Why would you ruin a perfectly good Linux kernel?

Re:CoLinux

[Anthony Liguori]

I noticed that whenever virtualization comes up, no one ever mentions CoLinux.

That's because no one really takes it seriously. It's not that it's not novel or not interesting, but the Linux guest runs as a *kernel thread*, not as a process. This means the guest has as much access to hardware as the host. It relies on a very well behaved Linux guest to not bring your system to a screaming halt.

If CoLinux ever adapts to run the Linux guest in a lesser ring (perhaps 1 or 2) then it will be considerably more interesting.

Re:CoLinux

[radarsat1]

Ah, thanks, I didn't know that. Interesting. Maybe there's no point after all, if virtualization is a better solution.

Re:Informative but you left out one important thin

[Anonymous Coward]

Actually, if you purchase one license of Windows Server 2003 R2 Enterprise you can run 4 virtual Windows Servers as long as they are on the same physical machine. They can be any version of Windows Server (2000, 2003, 2003 Enterprise). In our volume licensing agreement R2 Enterprise costs roughly 3 times what 2003 Standard costs. Since we are running alot of virtual machines it makes sense for us to purchase R2 Enterprise licenses. Our host OS is VMWare ESX Server.

Re:Psst. btw

[hotdiggitydawg]

...because combining the (relative) stability and security problems of Windows as the host O/S, with the (relative) user unfriendliness and reduced application market share of Linux as a host system is really the best of both worlds.

</sarcasm>

Seriously. Maybe for development, and/or if you're really doing it on the cheap. Otherwise I fail to see the benefits of such a setup.

Re:Psst. btw

[BadAnalogyGuy]

The big idea is that these are cheap, disposable machines that can be configured as you like. The obvious benefits like running multiple server operating systems are pretty limited, because as you say, you need to pony up the cash to really run something worthwhile in the first place.

But for development, you get a little virtual network to write your distributed apps. Or you get a little basic machine for some embedded Linux programming. Or maybe an on-demand alternative OS environment. For the home user running WinXP, there's no point in running another XP image or even a Windows Server image. But for the hobbyist, having the ability to run simultaneous sessions side by side is pretty cool and extends the project possibilities for anyone with a modicum of creativity.

That's not free!

[Anonymous Coward]

It requires a registration to get to. You're using the marketing-speak perversion of the word free. Free comes from freedom. That software isn't free to download nor is it free to redistribute.

Which is missing a QEMU trick

[ratboy666]

QEMU is not just an x86 emulator.

It is a system emulator. What it does very well is support Linux binary applications for other CPUs. Want to run an ARM binary on an x86? QEMU will do it. Want to run an x86 binary on a Sparc? QEMU will do it.

QEMU also does system level emulation.

As a special case, QEMU runs x86 on x86 as well.

VMWare and Xen don't do that.

Ratboy.

Re:Which is missing a QEMU trick

[Cytlid]

I've been working with VMWare and virtual servers for a while now (Xen still won't run on my main workstation at home, some ACPI problem or whatnot), but I was really amazed at QEMU. I never really tried it until I read this month's issue of LinuxJournal (all about Virtualization!) ... some of the Xen and VMWare stuff I was already familiar with.

QEMU's ability to emulate other CPUs is invaluable. You can emulate a MIPS architecture and test your favorite Linksys firmware (I believe the OpenWRT guys already do this). I would really like the m68k emulation to stabalize so I can run old Amiga stuff (or try linux on m68k). Or emulate an ARM processor , drop a PocketPC firmware on it, and test drive Windows Mobile software (or porting Linux to those devices). The possibilities are endless.

Re:Which is missing a QEMU trick

[Wolfrider]

Qemu does sound interesting. But the last time I tried it (admittedly months ago) it crashed while running a Knoppix Livecd that ran fine in Vmware. Has the stability of Qemu improved? (Honestly curious; I'm a VMware fan but have just downloaded Parallels and intend to eval it.)

Re:Which is missing a QEMU trick

[ratboy666]

QEMU is not a virtualizer. VMMware will do a better and faster job of that.

QEMU *can* be used as a virtualizer -- if you have a problem, report it.

QEMU can run x86 Linux on a Sparc (and so can BOCHS). Where they differ is that QEMU does so by translating the binary instructions. BOCHS has this available as a limited experimental feature, but generally interprets each instruction. Which means that BOCHS can run just about anything x86 *slowly*.

QEMU can run just about anything (x86, ARM, MIPS, etc.) on anything, and do so moderately (not as quick as VMWare, but many times faster than BOCHS does x86).

QEMU is built as a Linux on Linux, and as system level (hardware included) versions. The hardware (system level) is probably most mature for x86 PC, because it can leverage from BOCHS.

If you just want to run x86 on x86, VMMware is the way to go. If you want to debug or inspect tricky x86 sequences, BOCHS is a good choice. If you want to run foreign binaries, QEMU is the path.

Which is why I use all three of them.

The right tool for the job, and all that...

Ratboy

Re:Psst. btw

[adam1101]

> you have to register to download it

Direct links (copied from Digg):
32-bit
64-bit.

Overall Management

[Khan]

I think what makes VMware stand out from all of the rest is Virtual Center and what it brings to the table. Being able to manage ALL of your VM servers hosted on ALL of your ESX servers is a huge plus. And while this version of Virtual Center absolutely has is shortcomings, the next version of both VC and ESX are really going to raise the bar from what I've seen. The mainframe has come full circle.

Re:Overall Management

[jpvlsmv]

And what really makes VirtualCenter stand out is that VMware has bundled their VMotion product with it.

VMotion is the ability to migrate running virtual machines from one physical server to another. It's roughly the equivalent of Mosix for VMs.

And it works across ESX versions, so there is no VM downtime when we need to patch our VMware farm. And it works across (some) hardware platforms, so we can upgrade our VMware farm from 2-way servers to 8-way servers, again with no downtime.

--Joe

Performance on virtualized servers

[pnuema]

I'm a performance tester who has had to completely reinvent how we do business thanks to virtualization. How do you give assurances to an application that they will perform adequately in a virtual environment when by definition performance will always be dynamic?

The primary approach we have had to take was to stop looking at whether an app will perform on a virtual machine, and start looking at whether or not it will be cost effective for the app to perform virtually (in general, apps that will perform in the physical world can be made to perform in the virtual world if you throw enough resources at them).

It's an interesting problem. We found that our company's big push into virtualization had to be scaled back a bit - not every server is truly a good candidate for virtualization.

Re:Performance on virtualized servers

[Anonymous Coward]

Mod parent up!

Re:Performance on virtualized servers

[kma]

I'm a performance tester who has had to completely reinvent how we do business thanks to virtualization. How do you give assurances to an application that they will perform adequately in a virtual environment when by definition performance will always be dynamic?

VMware ESX Server provides proportional-share guarantees for CPU, memory, network and storage performance. I.e., if you always want 50% of a CPU, or 200% of 2 CPUs, or 75% of the bandwidth of a gigE nic, etc., that can be arranged.

HTH,
Keith (vmware employee)

Re:Performance on virtualized servers

[ovz_kir]

Well, say in OpenVZ you have near-zero performance overhead (and thus very-very-close-to-native performance -- in some situations you can not measure the overhead at all, in some others it is like 2-3%), so it won't do any harm to performance if you virtualize your real servers. From the performance side, each VE in OpenVZ has its set of limits and guarantees. And those guarantees will be met unless you "oversell". I.e. if you plan things right it will work right. And there are some benefits from the virtualization that you must not forget about. Some people, say, use OpenVZ to run a single virtual environment -- just because there is added value -- you can clone your VE, say, or migrate it to another box without a need to care much about hardware differences, etc. etc. Live migration is another piece of cake -- you can leave your apps running while doing hardware maintenance -- just migrate it to another physical server!

Re:Performance on virtualized servers

[Anonymous Coward]

We use VMWare ESX in our shop and it works great. It has reduced the footprint of legacy systems and low resource systems dramatically.

While it is true that you can guarantee resources to guest OS's, it still true that there will be systems that use such a large percentage of resources that it just doesn't make sense to VM them. Large database servers would likely fall into this category.

One thing to remember though, you are putting a lot of eggs in one basket. Include a back-up/recovery procedure in your design.

Like any other enterprise technology, it's not all or nothing, it's all about intelligent design and rolling out what makes sense.

David

Re:Performance on virtualized servers

[pnuema]

VMware ESX Server provides proportional-share guarantees for CPU, memory, network and storage performance. I.e., if you always want 50% of a CPU, or 200% of 2 CPUs, or 75% of the bandwidth of a gigE nic, etc., that can be arranged.

We are aware. The problem is, if I have to guarantee 2 CPUs to make an app perform, it is more cost effective to buy a physical box - those hosts aren't cheap. We determined our break even point to be 35% of a CPU - any more than that, and we make it a physical server.

The meaning of 'virtualization'

[Anonymous Coward]

Can't help noticing that when people say 'virtualization' these days, they tend to imply virtualized operating environments, virtual machines, etc., because that's what's hot right now. IOW there's a strong chance that 'virtualization' wouldn't imply, say, 'virtual file system,' 'virtual storage,' etc.

a noob must of wrote that article

[Anonymous Coward]

This sounded like a really short Xen advertisement. It's interesting how he mentioned the server infrastructure and really didn't touch any on vmware ESX. Granted I'm pretty biased to ESX server since it meets my needs and my customer's needs 98% of the time (the 2% being database servers which don't usually virtualize too well due to the IO issue), but overall all I really read about that stuck in that article was Xen. Would I trade a somewhat newcomer like Xen for a already developed product like ESX which sits in many environments..? Most likely not - although I wouldn't hesitate to put it in my lab as it would be great to see what feature it offers and it's approach to virtualization. The mention of low level system calls, sounds like that can be addressed by shares and affinity settings with ESX server to me on a per VM basis (but to what degree I'm not sure until I've played with both products). He didn't even go into mentioning how Xen would handle contention for a resource (considering he spent most his time talking about it). ESX and VirtualCenter (when properly deployed for what your trying to do) is a killer virtualization environment that I believe would be hard to beat - but I'm always trying to keep tabs on where technology is going and will welcome newcomers to the virtualization field. But that article was badly wrote, and biased in my opinion. If your going to tread into such a topic, alteast give each equal exposure. There seems to be a big emphasis on Novell despite the fact that I've barely ever come across a handful of environments that ran it. If your not a Novell sever shop, then VMware is where it's at *IMO.

too thin

[Anonymous Coward]

This article doesn't go into any detail whatsoever on the issues/drawbacks of virtualization. I was disappointed. I expected to see a little more info on why virtualization isn't like having a second computer because it shares all of the buses with all the other OS's. For many of us, CPU performance is not the issue, its I/O wait. Data intensive applications, such as those accessing databases or handling web traffic, don't benefit from virtualization if their IO bound and already having significant I/O wait on a single system.

I believe there are virtualization systems that overcome these kinds of issues through hardware architecture, but they aren't cheap. But sadly the article didn't survey any of them.

If you know what virtualization is, all this article does is introduce you to some of the software players. It doesn't even touch on the virtualization solutions from Sun, HP and IBM. I don't think its worthy of Slashdot top level visibility unless you only care about windows and linux and have little idea what virtualization is.

AutoSysAdmin

[Doc Ruby]

My dream is to run an installer on a single host on my LAN. Give it the root passwords of each of the various hosts on my LAN. Then watch as it halts and backs up each of those hosts, then installs the virtualization SW on each host, making a pool, then reinstalls each host as a virtual instance in the pool. Then I'd like to see the virtual pool balance load and failover among my hosts. When I add a new host to the LAN, I'd like the pool to just use the new capacity proportionately.

Of course a live, good sysadmin will be much better at that process until AI is more than just SF. But how close is an even barely adequate version?

Re:AutoSysAdmin

[tinkertim]

I've got this *kinda* done. This was our goal:

* Total centralized command
* Dynamically provision / reprovision based on application demand, roles and rule sets
* paranoid sanity checks
* use, but don't force LVM. Make use of just conventional images.
* Integrate HPC's that can be dynamically reprovisioned.

This is done with Xen + openSSI. In a case like yours it would not be a conventional single system image, you'd be using a couple redundant wasabi style NAS's, which can also be built with Xen for additional failover, storing many images and running a few separate director nodes pulling from various places.

Drawbacks :

* Xen 3 is still very beta so we're using xen 2, which has a memory cap.
* You really get stuck with just using iscsi for it to be practical over copper
* Vlans get a bit confusing, but are accomplishable (i.e /29 on one end /32 on the other)
* You need to be really comfy patching the 2.6 kernel
* Adding Win2k3 into the mix for provisioning will require xen 3 and mass stability, which is in sight .. but still kind of far out on the horizon.
* Xen's balloon driver is still not quite perfected.
* Depending on what you push latency could be a small, but over-comeable issue.

It utilizes CVIP / HA-LVS and the loadleveling inherent in openSSI to accomplish this.

What we're doing doesn't exactly fit your need, but could with a little work. There are at least a dozen small companies that have been following and playing with Xen working on something nearly identical. What takes forever to account for is 'the better idiot'. There are just so many possible configurations.

If you want to tinker check into Xen + openSSI , there's a wealth of info regarding it and even some setup scripts for it in the works. Many people want the exact same thing, including me :)

Re:AutoSysAdmin

[Doc Ruby]

So when you go the rest of the way to completing your project, I'd be able to take my 4-host LAN (4 different P3/P4 motherboard/CPU/RAM/HD configuration hosts), add a 5th host, replace their internal IDE storage with iSCSI, and run your software to create a virtual pool with the extra host's capacity? Or allow a host to fail with immediate autorecovery? Does it have to be iSCSI - what kind of NAS do you support?

Are we really talking about a "compute RAID" (RAIH?) available sometime this year?

Re:AutoSysAdmin

[tinkertim]

Doesn't have to be iscsi that's just what we're using because its simple. As far as the netfs , anything you can put into the 2.6 kernel will work. We figured on a local gig-e network over copper which is why we're using iscsi for developing it .. but by no means is it built 'around' iscsi.

It could be made to work with just about anything. That's the other part of why it's taking so long to do. But yes, you could accomplish that.

Bear in mind the definition of high availability is no single point of failure. So if each machine had just 1 physical nic, though you could simulate 2 .. you still have that bottleneck and potential point of failure. Your power supply is also a single point of failure. If you just want basic redundancy than what you have will work as long as you have gig-e nics and a switch supporting it.

Any time you are working off a SAN you may also want to contemplate using a local disk (SATA or SCSI) for local swap depending on the applications and how much memory you're willing to give them. That just makes sense.

I also really can't call it 'our' software. We did not write openSSI nor Xen .. all we're doing is putting them together and simplifying management of both, then combining them in a way thats practical to do what you want. Most of it will be open sourced once its stable enough to release so we can get more developers working on it.

A RAIH is available now, with the pros and cons I listed. What's lacking is the ultra neat spiffy cool idiot proof management console. When we release you can bet some very knowledgable skeptics are going to aim to tear it to shreds.. I mean thats a guarantee. So we're trying to do that ourselves in the development stage as we go, even if it means some major drop-back-and-punts.

You could do want you want right now with just the Xen SSI kernels , Debian and a knowledge of how to play with CVIP and HA-LVS. Once you grasp those you'll see how easy it would be to script what you want. I don't think it will be long until you see something (or many somethings) like this hitting the commercial market.

We plan to go the open source route as custom configurations would pay for and allow us to profit from the time / money that's gone into it (well, we hope, anyway) .. but just can't see charging for what's 80% open source to begin with. That's a little too close to squatting.

Re:AutoSysAdmin

[Doc Ruby]

Well that's all great news. I don't actually require "High Availability", any higher than a P4/3GHz/2GB/500GB (before it goes down :) in the pool. If I can pool the heterogenous IDEs in each host into a pooled RAID, that virtual server does what I want.

Now, if I can install it as an autoupgrade to my current (nonvirtual) LAN install, like I described in my original post, then it sounds like my dream is actually coming true :).

Re: P2V clarification

[jjohn]

For instance, compare VMWare's P2V application to convert Physical to Virtual servers against Microsoft's offering which requires having a spare server lying around which must run Windows Server 2003 Enterprise with Automated Deployment Services.

I hope you will endulge this spam-vertainment.

For the record, there are third party vendors of P2V software including Platespin and Leostream, whom I work for.

The VMWare P2V Assistant is arguably easier to use that Microsoft's VSMT solution, which appears to be more of an API than a standalone application. Of course, these vendors only support virtualization to their own platform.

Platespin's solution is pretty robust and feature rich, if a bit on the costly side. Coming in as the cheapest and caveman-simplist product is Leostream's P2V Wizard, which supports MSVS, ESX, GSX, VMware Workstation. Leostream also offers free trials so that you can see if the Wizard will work with your hardware.

P2V is a weird process. In the real world, if you shoved a hard drive with an OS from one machine into another machine with complete hardware, you would expect to spend a lot of time replacing drivers. The P2V products all attempt to do this driver replace automatically (since the target virtual machines have a limited set of hardware, this task is managable). Similarly, V2V is pretty simple (even more so than P2V). V2P is very tough because you have to support a large universe of hardware, which Microsoft and Linux OS engineers will tell you is no fun at all.

Virtualization isn't an appropriate solution for all classes of applications (I'm looking at you, heavily-used Oracle DB), but for many tasks VMs make system administration a magnitude order easier. VMs are outstanding for creating standard dev and QA environments.

As for picking a virtualization platform, you need to define what you want. The best performing virtualization layer I know of today for the x86 architecture is VMWare's ESX. It's been in production for years now and it shows. The console OS of ESX is a modified RH7.x system, so linux hackers will feel at home (Windows guys -- not so much).

GSX and MSVS are just about even, offering decent, if not stellar performance. I have a hard time recommending GSX, since it's the middle product between the high-performance ESX and the low-end Workstation.

Xen is a newcomer. I haven't use the newest Xen which, with the right dual-core CPU, will support W32 guest OSes. It's a work in progress, so it's probably not what you want to slam into a production environment. However, it seems to have a good architecture (using a hypervisor model), so I'd keep an eye on it.

Then there's Parallels, which looks a lot like VMware workstation. It too is a work in progress. I think Parallels could challenge Workstation on W32 and linux, if development continues on that product. It will be a low-cost alternative to Workstation. Eventually, they will need an enterprise server product to go against ESX.

These are the hardcore virtualization options. There are *partitioning* schemes, as used by UML and Virtuoso, which sort of work like VMs. These don't create VMs, but separate instances of the underlying OS. I haven't worked with either of these very much.

Whatever you do, consider spending an afternoon getting to know *some* VM technology. You'll be glad you did.

Do a little more digging...

[mrscott]

I manage a fairly small (40 server, all Windows) shop and, while I technically like VMware better, I can't do a cost justification for it (I've tried). Another factor to consider in the cost equation: If you use VS 2005 R2 and opt to use Server 2003 R2 Enterprise as the host, MS allows you to run four virtual instances of Windows Server R2 on that hardware for no additional charge.

Yeah, ESX is better in some ways, and VMware's tools are better than MS's. But, VS 2005 R2 is actually a really good product. We're planning our entire virtualization strategy around it and have been very pleased with it thus far. And the licensing changes have made it very attractive.

Worthless

[pooh666]

There is NO WAY to write a *short* article on this topic without being a total waste of everyone's time. The sentence that wraped up Zones in the same space as Vservers was misleading and one of several points that didn't have the slightest chance of being explored. But as always on /. the comments will make up the difference. No wonder no one RTFA..

meh... TFA is worthless

[Jay]

It only talks about virtualization as it exists on Intel hardware? What about Power LPAR? What about Solaris Zones? What about mainframes? How can it be called an overview when it only touches on a third of whats out there?

I'd guess cause thats what the author could get for free in the 10 minutes it took to write that article.

Re:meh... TFA is worthless

[ovz_kir]

> What about Solaris Zones?

He actually mentions Zones: Vendors following single OS image approach include Virtuozo, Vservers and Zones. This method groups user processes into resource containers and manages access to physical resources. While this approach can scale well, it is hard to get strong isolation among the different containers.

The problem with the article, as I see it, is that he claims something without supporting his words, and the quote above is the excellent example. Why he thinks that it is hard to get strong isolation in OS-level virtualization is beyond my imagination. Consider Virtuozzo which is used on thousands of boxes and let ISPs sell cheap virtual environments VE with root access -- they would not be able to use Virtuozzo in case isolation is weak and VE root can do something evil to the system. The fact it, hosting is a very hostile environment, and almost every major ISP uses Virtuozzo or OpenVZ -- that means it works and isolation is strong enough.

Re:meh... TFA is worthless

[h4ck7h3p14n37]

Why is it that you only get mod points on days when there are no articles you're really interested in?

I agree that this article was next to worthless. It basically listed the names of some virtualization technologies, and then said go with Virtual PC or VMWare. I too was looking for an overview of the virtualization concept with perhaps a discussion of how X technology differs from Y.

The company I work for just purchased a half-dozen p505 servers, a half-dozen p550 servers and an HMC for some new B2C websites. I was really hoping for a discussion of LPARs versus Zones (our main B2B site runs on UltraSparcs and is being upgraded to Solaris 10).

Slashvertisement

[frankie]

Jane Walker == TechTarget

Seriously, just two minutes on Google led to MUCH better articles, e.g. CMPNet, eWeek, and Virtualization.info

Mailing it in.

[Sergeant Beavis]

If Paul Ferrill is collecting a paycheck for writing this, then he should give it back. The article is a montage of nothingness. Did Ferrill actually do any research on the subject? It seems to me that he must have read some Novell marketing material but little else.

Ferill mentions, "On the downside, the x86 architecture does not lend itself to efficient virtualization." however he appears oblivious to Intel VT or AMD's Pacifica chips which are made specifically for virtualization.

Lets not forget that Paul doesn't even dedicate a sentence to one of the most important Virtualization products on the market. Hey Paul, go Google the letters ESX maybe you'll actually learn something.

VMWare NOT Linux

[spitek]

"If you want to use Linux as your host OS, you'll definitely have to go with VMware."

What is he talking about? From this one sentence it is obvious that this person should have never written this article. VMWare ESX server is not Linux. VMWare ESX VMWare's own creation kernel and all. Not to say they could have barrowed parts from who knows where but it is not a 2.4.x Linux kernel or anything of that nature.

Now to the inexperienced user, installing and administration VMWare ESX you might get the impression that VMWare ESX is linux since the "service console" is based off of Red Hat. But this is not the hart of ESX, merely to local and remote configuration interface into ESX.

Re:VMWare NOT Linux

[carlcub]

ESX isn't the only flavor of VMware. There's VMware Workstation, as well as the new (free) VMware Server, which do, in fact, use Linux (or Windows) as the host.

Drawbacks? Pish posh!

[eno2001]

I'm using Xen on Gentoo Linux running on an OLD Celeron 400Mhz box with 384 megs of RAM and it runs three VMs at completely native speeds. There ARE no drawbacks. Period. Xen blows EVERYTHING else away in terms of ease of use, flexibility, and even the ability to keep a VM running while it's original physical host is down by migrating it to another physical host. When you combine it with Vanderpool or AMD's upcoming Pacifica hardware virtualization techniques, the sky will be the limit.

Re:previous hot topic: virtual reality

[utlemming]

I do exactly that. I run virtual machines to make sure that the computer runs well. I use VMware for the main reason that it offers some serious benefits, like support, virtual lan segments, and its flexability.

One of the main reason that I use VM's is that it allows me to dink around with malicious programs with hosing the host. I have even found that I don't worry if I install a malicious program or even for that matter a suspect program since I can rollback or run it under a non-persistant disk. I have a different VM for each set of programs that I run. For example, I have an Oracle environment for developing Oracle programs (I know, not the brightest idea, but I can run Oracle when I need to without breaking it with another program), a work, school, a playgrouind(or sandbox), and even a personal Windows enviroment, and various Linux distributions. My wife even has her own VM. It has proven a stable way to enjoy and play around with programs an not have to worry about a lot of things.

Re:Psst. btw

[Poltras]

You know there is no 3D capable hardware emulated under VMWare (neither any other solution), which makes visualisation of 3D (or even 2D, since the hardware emulated is not capable of accelerated 2D sprites and other stuff) graphics really slow? And why the hell would you play UT200x under a VM anyway? It was not meant this way. Buy an old P4 at 300 bucks and you've paid your VM license.

Re:Psst. btw

[Anonymous Coward]

Games aren't the only reason for 3d stuff (and fancy 2d stuff), for example scientific apps may have visualization components that use 3d models.

Altiris Software Virtualization

[daHIFI]

I'm surprised the article didn't mention software virtualization. The package from Altiris creates a virtual layer between the OS and the file system. It captures changes during the setup and then creates an virtual application 'layer' that you can turn on and off at will, or reset back to a default state.

The way they're pitching it to my department is that we can use it to deploy applications Enterprise wide with little testing. If a service pack or patch breaks the system it can be turned off. If a user deletes a crucial application file or an app crashes we can reset the layer to a working state. It also allows you to have different versions of a program on your system without causing conflicts.

I am using it for app testing. I can install an app in a virtual layer, give it approval or not, and then disable the layer and it's as if it was never on my system to begin with.

It works great so far although there are some problems making application layers. For example once you install Firefox in this way and download something it wants to capture the download as part of the Firefox layer and if you turn off the FF layer the download dissapears also. This kind of application specific tuning is the worst part about it but they have a nice user group page setup @ http://juice.altiris.com/.

The software has also been released for free non-commercial use and can be gotten at Download.com

Re:Altiris Software Virtualization

[Feebleminded_Genius]

I've been testing this recently as well.

A solid product, though I dislike the fact that the application layers are stored locally. When you step back and look at it, all that you're really getting with this product is the ability to create a custumized install package that can repair itself, which in my mind sounds remarkably similar to an MSI package and Windows installer. The only difference really is that Altiris SVS (I believe it's called) provides a more reliable uninstall method and gives you more control over repairs.

When you have a moment, take a look at the solution from Softricity. We recently had a vendor demo of the product, and it really is the more advanced of the 2 tools and is probably a better fit for larger enterprises. It stores the layers on the network, and provides the ability to download only the necessary components of the application locally. Their online demo shows MS Word running with something around 30% of the application loaded locally. Interesting stuff, I'm looking forward to getting my hands on it.

Either way, you are correct, software virtualization is coming. I look forward to the day when we can deploy PC's to our users with nothing but a clean image, and knowing that it will stay clean.

What a bunch of ludicrous claims.

[Anonymous Coward]

The only thing I'll say is that you are clueless. It's very obvious you haven't even used virtualization software. If you had, you wouldn't be making these ludicrous claims.

Since other people have explained just a few of the reasons why you're such a dork, I'll leave my comments to what I've already said.

Re:previous hot topic: virtual reality

[Ohreally_factor]

Does that four-way dual core Opteron set up include a cloaking device?

Is Corporate Speak Invading Your IT Department?

[umedia]

No, but's invading this thread and my inbox. So help me if I get one more e-mail with one more product that adds Virtualization as if it's the new iPod, I'm going to get cranky...

Re:Drawbacks? Pish posh!

[ovz_kir]

If you are so happy with Xen, I suggest you try OpenVZ (http://openvz.org/ -- I bet you'll be even more happy. Unlike Xen, OpenVZ does not have that big I/O overhead (our tests shows Xen guests do I/O about 30% slower than native system). The biggest thing though is you can run not 3 but 30 virtual environments, and dynamically manage their resources (like adding/removing memory from the environment without any need to restart it).

Finally, live migration for OpenVZ will be released Real Soon Now.

Re: P2V clarification

[k8to]

P2V is a weird process. In the real world, if you shoved a hard drive with an OS from one machine into another machine with complete hardware, you would expect to spend a lot of time replacing drivers. The P2V products all attempt to do this driver replace automatically (since the target virtual machines have a limited set of hardware, this task is managable). Similarly, V2V is pretty simple (even more so than P2V). V2P is very tough because you have to support a large universe of hardware, which Microsoft and Linux OS engineers will tell you is no fun at all.

I disagree. I've done this many times and in the real world I have expected this to just work. In old Linux systems, it would boot up, I would make a few changes, and then everything would be fine. In modern Linux systems, everything is autodetected and fixed for me. In BeOS, the drive image booted on the new computer completely correctly without a single notice or warning with full functionality. On mac systems, the drive booted without notice or warning or problem. Dos of course always worked.

The only system which has given me significant trouble of this nature is Windows. It is astounding just how painful it is. So given that the meat of these products are pointed at windows, I agree with your message, if not your text.

Re:Psst. btw

[nolife]

I have a VERY premature assessment of the the MS Virtual server product.
To be fair, I have only been using it for 2 days.
I don't like it ;). I do not even have a virtual server built on it yet but I am getting there. I created a 12GB virtual disk and I am using a Windows 2003 Server iso mounted on a virtual CD to do a fresh install of the OS. When I left work today, It just got done formatting the 12GB partition but that took over 2 hours. That is like the third step in the OS install. Another thing I do not like is the ActiveX control for controlling the virtual host. It seems and there is no way to bounce around between the server controls and the virtual host without closing the remote session and opening it back up again. I think a standalone application would be much better, maybe they have one already.

For a comparision, last week I installed VMWare server (the free version) and had it up and running in no time, and had a virtual server installed in less then 60 minutes from scratch. We have been using ESX for years so maybe some of the lingo and layout is already familiar to me. Another bonus with the VMWare Server is the network drivers and SCSI disk drivers are the same as ESX and our dos network boot disk works fine. We use the gold disk concept for all of our workstations and servers. I can boot a new virtual machine with our network boot disk iso, connect to our share where GHOST.EXE resides and pull our standard server image over just as you would do for a physical server. With VMServer and ESX, I can install that ghost image in about 20 minutes. I modified our network boot iso to add the driver that the MS Virtual presents to its virtuals (A DEC/Intel 211XX) and although the new virtual boot image booted fine and connected to the network, I was getting less then 100K/sec pulling the image and Ghost was reporting some time over 24 hours to complete. Maybe the DOS capabilities of MS Virtual server is severly lacking (as noted by the slow install and the slow DOS performance running ghost). Who knows.

To sum it all up. I have very little face time with the MS Virtual server product, but I had very little time with VMWare Server as well. My ESX background may have made my VMServer experience better so who knows.

New Ipod server with Virtualization Technology!!!

[What me a Coward]

Buy now the new Ipod server with virtualization Technology

    You wont believe the number of songs you can serve to you're PC's in the home as well as ipods over 802.x wifi!!!

    Act now and get 20 free music downloads to you're new ipod server with Virtualization!!!

    You can't go wrong this is a must buy, It's Jobs approved for you're home!!!

    Act now don't wait for the next Virtual server virtualization Ipod Server that we'll be offering as a paid upgrade in 6 months time. You will be sorry and left out in the cold with extra costs to pay for the Ipod server virtualization server to enable the upgrade to the Virtual server virtualization Ipod Server if you don't act NOW!!!

    Click link to approve sale!! Don't wait click Now!!!

  HTTP://www.IpodVirtualserver.com/Virtualization.ht ml/

    Well your post was fairly screaming out for it :D

Re:previous hot topic: virtual reality

[Decker-Mage]

It will. Actually, I've had that for years here for my 'net access through two pieces of software. Unless I unblock/uncloak (and I can do it selectively) all you see in your website logs are a date/timestamp and an IP address and I can change that IP address on a second by second basis if I'm really feeling beligerent. I take my anonymity seriously when I'm spelunking some segments of the web (mostly keeping track of what the system crackers are up to for my security interests). Usually I'm not that serious about it.

Re:Psst. btw

[Decker-Mage]

It's something that VMWare is experimenting (playing) with and marked has "Experimental" for a while now. As for playing games, well that's one way to test it but there are quite a few apps out there, serious (e.g. scientific) apps, that also make use of the Direct3D functions.

When I do a beta for someone, I not only test the working stuff (which all gets certified first here) but also test all the experimental stuff as well. For instance, Solaris 10 is also "Experimental" and yep, I test it, each draw. Maybe that's why so many big firms keep asking me back.

SUSE

[Anonymous Coward]

SUSE (Novell actually), gives 1 million to Mono (instead of 100K to python, perl, PHP, mozilla, ruby).
It dropped KDE (although SUSE was a KDE distro), because Migule De Caza is a microsoft agent.
It promotes VMware instead of qemu.

And makes lots of money by selling open source software while it gives nothing back to the programmers. I therefore think that Novell is worse than microsoft. At least microsoft sells their own software.

Re:Drawbacks? Pish posh!

[eno2001]

Thanks for the interesting link. I'll probably take a look at this for testing with a specific application. However, reading the FAQ the only thing that seems to be missing that Xen does provide is controlled directe hardware access. For example, if I've got two video capture cards in my Xen system, I can assign each of them to their own specific VM. That way I can have two VMs acting like independent PVRs. Similar things can be done with NICs if I want greater isolation between physical networks. But I see that OpenVZ definitely would fit well in a server room environment for SQL DBs, web, mail, DNS or other network services.

Re:Drawbacks? Pish posh!

[ovz_kir]

No probs, here is the extract from vzctl(8) man page:

       Device access management

       --devnodes device:r|w|rw|none
           Give access (r - read, w - write, rw - read write, none - no
           access) to special file /dev/device from VPS.

Re:Drawbacks? Pish posh!

[ovz_kir]

And in case you want to have a dedicated network device in an OpenVZ VE, use something like
vzctl set VEID --netdev_add ethX --save

Re:Drawbacks? Pish posh!

[eno2001]

Thanks again. I'll definitely have to give this a look over to see if it fits my needs and how it runs on my admittedly old hardware.

Summary of TFA

[Sigg3.net]

Virtualization... proliferation... underutilization... isolation...

It's amazing!