Slashdot Mirror
Longhorn Server Will Stress Virtualization
Rob writes in with an article from CBROnline based on an interview with Microsoft's UK server director. He says the timing of the release of the next version of Microsoft's server OS, dubbed Longhorn, depends on the company getting virtualization ready to go. Microsoft has apparently decided to embed its hypervisor technology into Windows, an OS-centric approach to virtualization shared by XenSource Inc., its open-source rival and partner. This contrasts with the model of virtualizing the hardware layer being pursued by VMWare. The Microsoft spokesman is coy about a release date for Longhorn, saying it could be earlier or it could be later (but it should be in 2007).
Is it Long and Horny, or is it Micro and Soft?
...and one vmware-server-linux to try and delete them all.
This will work fine if all the servers you want to run on a given machine are MS. I like VMWare for the fact that you can load Windows on one VM, Linux on another, and Solaris on yet another. The folks at PACCAR are running massive numbers of systems on a single Blade.
What I would like MS to give us is a Virtual Platform OS, much like VMWare's ESX server. Give me an extremely lightweight OS geard towards Virtualizing the HW layer, then let me load anything into each VM.
As far as I know, VMWare is the only one doing that.
Politics is the art of looking for trouble, finding it everywhere, diagnosing it incorrectly and applying the wrong fix.
wasn't it?
Visit http://ringbreak.dnd.utwente.nl/~mrjb/growingbettersoftware to download your free copy of the book
Is the virtualization gonna be recursive?
If so, how will they handle licenses of licenses of licenses of...?
And will Active Directory be able to handle trees of trees of trees of... license keys?
Microsoft seems to be TERRIFIED of VMware (EMC). Why? Is is because VMware allows the use of Windows UNDER Linux? (Whereas, it seems, Microsofts preferred approach will be Linux under Windows?).
Of course, having Linux as the HOST OS means that driver vendors will have to support Linux more in the enterprise. And, I believe that Linux is the "better" OS in that the kernel has gone through a more stringent review process.
But all of these statements -- that the Windows Virtualization Technology will be stunning, that Virtualization belongs in the OS, etc. seems to be thowing FUD directly at VMware (and, I assure you, the VMware product is "stunning" -- I particularly like the Server product running on Linux).
To my knowledge (or my opinion, if you prefer), Microsoft ONLY reacts this strongly if their platform is being threatened. And I don't see what the introduction of a bit more enterprise driver support does to threaten Windows.
Ok, I have a guess: It could be that Linux is so good that it makes for a more solid OS base. This then provides a compelling platform to begin virtualizing Windows Servers on. But, if the enterprise is ALREADY basing on Linux, why not start transitioning to native Linux? After all, its stable enough to host Windows, right? A chink in the platform.
But that implies that Microsoft believes that Linux is that good...
Just another "Cubible(sic) Joe" 2 17 3061
The amazing thing is that no one's tagged this "defectivebydesign". I know you all want to.
talking about getting it right? so the degree will be
Yawn...if Microsoft could write an OS that had decent multi tasking, a responsive scheduler, and adequate memory allocation and protection, the appeal of running a bunch of virtual machines just to run a bunch of different jobs and keep them from interfering with each other would be much less. This has been done before, most notably by VMS..I used to manage a cluster of large VMS systems, each of which had dozens of Oracle databasea on them, supported interactive editing of documents for hundreds of people, and ran a mixed bag of financial, accounting, engineering and program development applications...all on the same machines....looks like Microsoft and Cutlerdidn't incorporate enough of it in Windows...
There is no God, and Dirac is his prophet.
*NM*
Where did this story come from? 2005?
GArtner predicted that viriutalization would be a big factor in a next version of the next major OS.
I'm virtually certain that this will come with virtual customer service. Now if only all the bugs would be virtual and I could pay with virtual cash, I might be more than just a virtual customer.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
I'm fairly sure the hypervisor part of the virtualization-stack will become a commodity... (Though i prefer the ESX-model, with a dedicated, stripped down virtualization host).
The real business in virtualization will be management tools, - I bet VirtualCenter/Lab Manager will be VMware's main business in a few years.
I'm pretty sure Longhorn Server will stress just about everyone who comes into contact with it.
Because Windows machines need to be rebooted all the time for bugfixes, application death, and patches, MS wants you to be able to run several instances of Windows on 1 piece of gear. A decent virtualization layer will allow you to migrate users from 1 instance to another, so you can bounce the instances one at a time. That being said, VMWare currently totally rocks at this, so MS has an uphill battle. One which they'll win, of course, because they have the mindshare and the Windrones will buy it.
I want to delete my account but Slashdot doesn't allow it.
They basically gave up pretending Windows is a multitasking multiuser platform and now start recommending one Windows per one service. This is of course what everyone has been doing since Windows servers started getting deployed. It's HELL to keep windows with one service running operational, because the system is a black box of maggots. This used to sell lots of server hardware.
Longhorn on the bottom virtualization enabled, n longhorns on top in sandboxes, guess whether the suc^h^h^hclients have to buy one or one+n licenses?
Each one of those sublicenses being licensed server 2003-style? Pay more for each connection?
One server dedicated for Virtualization Interoperability Manager 2007 Pro Signature Version? (a la TS client service?)
They lost, they know they lost, now the only way to keep their marketshare is legislation and DRM to keep the format lock-ins and infrastructure lock-ins in place.
'Once scientists, even the dim-witted social scientists, get muzzled, the Western Civilization is finished.' - oldhack
Two ducks walk into a bar
One duck asks the other, "What time is it?"
"Time to drink!" he replies.
-- www.globaltics.net
Political discussion for a new world
From the looks of things, it'll put stress real hardware as well.
"Have you ever thought about just turning off the TV, sitting down with your kids, and hitting them?"
ESX 3.x management client requires a .NET platform running on windows. No mono, no wine. Yes, it's snappier than the web interface, but jesus - they should bundle as many free windows licenses as the client requires with every ESX3 sold. It's BS I have to run VMware 3 Infrastructure Console in XP in VMware Workstation on Linux. That's one winblows license for no extra functionality and tons of RAM and resources wasted for this ludicrous tie-in.
To rephrase: they sell a lean and mean proprietary VM hypervisor kernel that uses linux for management and stuff. It can run on any OS. And you're required to run a closed proprietary OS to manage it.
This is not only insane it's DANGEROUS. What if M$ broke .NET in the next hotfix so that VMware ESX 3 management software broke?
There have been demands for a mono or unix or linux native client to manage ESX3 for at least 18 months and STILL no official word from VMware. I wonder how much money M$ paid VMware to get one of their worst competitors to bend over.
'Once scientists, even the dim-witted social scientists, get muzzled, the Western Civilization is finished.' - oldhack
Awesome! Here's how it will almost be implemented.
Beta 1 will have it. It will totally destabilize the OS.
Beta 2 will not have it, but it will be replaced with shiny graphics.
Beta 2.5 will have to remove the shiny graphics, because these too will destabilize the system.
Beta 3 will put it back, working perfectly.
Rc 1 will be totally unstable and also have gaping security holes.
Rc 2 will look like Server 2000
Rc 3 will look like "longhorn" but without the virtualization. However, the shiny graphics will be there.
Anyone signing up for the "upgrade"? I hope you like vapour.
blah blah blah
VMWare's hypervisor is capable of using VT and Virtualization capabilities of BIOS/HW to improve performance, provide more features. Xen however is near native speed and doesn't require VT or Virtualization to run, but is enhanced by VT and Virtualization. VT and Virtualization for Xen allows FV machines to run "un-modified" OS. Longhorn will be modified it appears to run in a Xen container (para-virtual). This will be interesting to say the least. The question is will I be able to run modified Xen OS in a Microsoft Virtualization Platform? Can I run RHEL5/Virt or SLES10/Virt on a Lonhorn server? If so, cool. If not, why not? Put simply, will Longhorn be a Dom0 and can I run RHEL/SLES/BSD/Solaris domUs. As well, can Longhorn be a domU running on a RHEL5/SLES/XenEnterprise Dom0?
I will take a Hardware solution over a Software solution (remember those fun-assed Winmodems?) any day of the week.
Microsoft's Virtual server sucked ass. I see that they plan on carrying on that fine tradition into the future.
Given Microsoft's penchant for taking a technology and re-branding it (i.e., adding security problems with one-off "standards"), is anyone else afraid of the security implications of a MS version of virtualization?
I don't think MS has any SERIOUS interest in virtualization technology other than the fact that it increases their licensing revenues. Why are the re-inventing the wheel? Are they just trying to take control over it to keep tabs on which installations are legal and which are not?
Maybe I'm just a skeptic.
I might know what I'm talkin' about, but then again, this is Slashdot...
This is great news! I'll be sure to put it on pre-order in 2009.
Or I could wait for some 12-year old in Russia to offer to sell it to me. I hear it comes with some pills to increase my m4nh00|).
Just wait until the VMware guys get DirectX 9 working at speed. Then move that to other operating systems.
Imagine being able to play your Windows games, but on an OSX box. Or Linux someday. It would be fantastic. Just make a VM, install your XP on that...then the game. And disconnect the virtual network card so your VM doesn't get pwned.
Weaselmancer
rediculous.
"I'm fairly sure the hypervisor part of the virtualization-stack will become a commodity..."
In a way it already is.
Why is there all this interest in virtualization?
Is it because it isolates the services from each other, because people can't write services that don't trash their environment and thereby corrupt other services and/or the OS? If so, that seems to be both the service's fault for being so badly written, and the OS's fault for doing such a poor job of protecting services from each other.
Is it because Windows can't multitask well and/or doesn't protect processes from each other well? If so, why does anyone think that another layer of Windows is the answer? If Windows can't protect processes from each other, why does anyone think it can protect VMs from each other? If it can't multitask well among processes, why does anyone think it can multitask well among VMs?
In short, why does anyone think this is the answer? Isn't the answer to get a real OS, one that actually works?
There are alternatives to the hypervisor.
From what I remember reading, the Windows Hypervisor will be a mini-OS below all VMs, including the host OS. But there has to be one designated VM to act as host OS, because it'll be the one responsible to host all drivers, that'll be accessible for the guest OSes via sort of proxy methods using the hypervisor (mainly only for the Windows based guest systems, unless other systems are going to implement the interfaces and functionality).
If you're running your Win32/DX games on OSX, then it's an OSX machine that happens to be running Windows as a task or translation layer. The computer is not a Windows box.
MS is all about market share. Without that, they're nothing. That's why they perform stranglehold tactics on PC manufacturers, like this. If people can run to the store and buy a piece of software and run it anywhere, then what's the point of Windows?
Most of us already own an XP disc. With no reason to buy another one, the whole Windows revenue stream dries up.
Weaselmancer
rediculous.
VMware does already have a modified Linux kernel available that includes that I believe is similar to what Microsoft is proposing for Windows. One can only hope that Microsoft works with VMware for their Longhorn release so that what is included works in conjunction with VMware and their VMI specification
I run a small VMware shop and one of the added benefits for us has been that we are able to corral Windows rather then extend our use of it. ESX Server 3.0.1 is very stable, we can patch and reboot on our terms. Windows in general has proven to be far less reliable, so running Windows VMs on Windows would just compound our problems. Sure, MS will stress virtulization in Longhord, but will jaded admins like myself do the same? I think not.
Its interesting to observe the way that Microsoft have always really fought hard to guard the who-runs-who position. They've always tried hard to "embrace and extend" (read: kill) any independent software from running windows OS, but are fine with Windows OS being the top layer and running it. Very wierd and this is just the latest move in that long battle.
I think it could be both Micro and Horny or Long and Soft....
nothing about these are exclusive, except you can't be SoftHorn, or rather Soft and Horny.
Get your Unix fortune now!
The issue is not one server per OS, it is one customer per OS instance in a hosted environment. Linux would do the same thing, after all just because it can have multiple regular users doesn't mean that customers who want root access will be willing to share one system. The virtualization allows this.
AntiFA: An abbreviation for Anti First Amendment.
I've read a few comments about virtualization making disaster recovery less painful (thanks to hardware abstraction) and server fault tolerance (due to redundant SAN attached boxes handing server images back and forth). That sounds great but let me explain how it will really be sold to CEOs and CFOs:
HARDWARE SAVINGS. I can already see the Microsoft ads....reduce the size of your datacenter, save electricity, buy less hardware...all by putting multiple mission critical servers on ONE physical server! That's why you need Longhorn server!
And that's how management will buy and implement this thing. All your mission critical software eggs in one hardware basket.
-ted
Is that the same reason IBM added virtualization to their mainframe OSes 30+ years ago? Because mainframes with one service are so hard to keep running?
Or is it that Linux and mainframes have perfectly good reasons to support virtualization (consolidation, deployment flexibility, disaster recovery, fault tolerance, etc.), while MS supports virtualization only because Windows is crap?
dom
That MS want to be DOM-0 (the host VM) which makes zero sense for anyone using virtualization. Windows licenses are per CPU, that means one Windows license lets you run as many Windows guests as possible sandboxed under a linux host. Microsoft don't like this and they're already positioning for the same protectionist, anti-competitive play we're accustomed to. This is exactly why we have competition law, so that even a monopoly supplying a commodity product be subject to market forces.
It's only a matter of time until their EULA is challenged in court, leaving Microsoft dead in the water.
Software is sold as a stack from operating systems through middleware to application programs. Software companies have to work with the fact that customers often buy a mixed stack from different vendors. Software companies spend a lot of thought on their sales strategy within the stack. They often see sales opportunities elsewhere within the stack as for example Red Hat recently went up the stack by adding Jboss to their product line. Software companies also worry about competing companies which are positioned below them in the stack somehow forcing competitors in the upper portions of the stack into a competitive disadvantage. When Red Hat began offering Jboss in competition to Oracle Oracle countered by offering support for Red Hat Linux mostly as a defense to prevent Red Hat from biasing Red Hat Linux to favor Jboss over Oracle.
I think that worries about competition within the stack played a part in Microsoft's decision to go down the stack and enter the virtualization market. Microsoft is probably worried that Vmware or Xen could place Windows at a disadvantage in a virtualized machine any time that they want. Also if Microsoft eventually manages to dominate the virtualization market then they could begin introducing biases against competing operating systems.
---------------------
Steve Stites