Robert McLaws: Windows Edition

Blogging about Windows since before Vista became a bad word

Microsoft's Virtualization Strategy is Doomed To Fail

I don't think Microsoft is "in it to win it" with Virtualization anymore. Mike Neil, GM of "Virtualization Strategy" announced today that Microsoft isn't putting out a beta of Windows Server Virtualization until Longhorn Server RTMs (which is in November, the last I heard). On top of that, they're going to be cutting some of the features they touted the most in previous public demonstrations; features such as live migration, hot-adding resources, and support for extreme multicore.

Now, normally this wouldn't be a huge deal, but it shows a larger pattern of malaise from the Virtualization group at Microsoft. As I understand it, Microsoft has outsourced offshored the entire development of both Virtual PC and Virtual Server to India, and Microsoft has all but abandoned future versions of that platform. In case that I wasn't clear, that means that those products will be in "servicing" mode (security bugs only) after Virtual Server 2005 R2 SP1 RTMs. This accounts for the fact that Virtual Server 2005 has been itself extensively delayed, and both platforms have not been advanced enough to compete with VMWare. While at one point it seemed that Microsoft was competitive, now they're being left in the dust.

(Offshoring it is not a bad thing, as great stuff comes out of Microsoft's operations in the East. But the core Connectix team is not working on VPC/VS anymore, and as of right now new versions will be bug fixes only.)

This "resources" thing is a load of crap. If Microsoft really saw virtualization as "the next frontier", they would bring all their weight to bear on the market, and blow the competition away. Instead, they're satisfied to outsource and kill their existing products and cut features to meet a deadline that was apparently too ambitious for the team. Actions are clearly speaking louder than words here, and Microsoft is headed for a really embarrassing loss.

There was a point when I was really excited about what Microsoft was doing with Virtualization. I'm not anymore. I can't afford to wait another year for a virtualization platform that's been delayed and castrated. I was just getting ready to deploy a new web server network on Virtual Server, in anticipation of WSV. Looks like I'm going to have to look into VMWare instead. XenSource is shipping now, maybe I should take a look at their platform too.

I had such high hopes. <sigh> Oh well.



  • H L said:

    Lets just get one thing clear... you have no idea what you are talking about.

    1. The timelines for Viridian are on schedule ... check out all the announcements since WinHec2006.

    2. Even VMWare hasn't put much backbone into their Windows host based product. Its only enough to compete with Virtual Server 2005 R2. Like M$... their bread and butter is the think hypervisor model.

    3. As for resources... Microsoft will release a virtual machane manager thats on par with VirtualCenter mid this year... and an update in lockstep with the hypervisor release. The hypervisor itself is a brand new undertaking for Microsoft. 18 months ago everything was just on paper and by January of 2008 they will have an enterprise ready management tool (in version 2) and a hypervisor that is probably 90% of what VMWare does today and for the next 8-10 months.

    4. As for resources... if you have ever coded in a trading environment or software dev group... you will know that your timelines are ALWAYS dictated by your feature list. The more features... the more you have to test... the more you test... the more time it takes to release. Its a matter of resources. Writting a lot of code is only part of the task... testing it... at a scale like M$ and VMware is no simply task.

    Yes... it sucks to lose live migration. Extreme multi-core? hehe... show me a tier 1 Intel/AMD server vendor with more than 16 cores in a box (btw... you seem to have clue... so you know the timelines for quad core ... show me the timelines for >4 core... 8 core), and the hot add of resources sucks to lose as well... but of the three you know live migration is the hard pill to swallow.

    May 10, 2007 2:42 PM
  • Microsoft hasn't 'outsourced' it - there is a team at the India Development Center (where I work too).

    May 10, 2007 3:49 PM
  • If you can't do something right, then don't do it all. The Windows Server Virtualization team is having

    May 10, 2007 5:13 PM
  • Josh Bernard said:

    I wonder if the delays are in any way due to MS building support into Windows Core. AFAIK Windows Core supports only very basic server roles and to include something like virtualization would require alot of work in this area. To me having virtualization in Windows Core would be the biggest feature to have. With virtual server you are running a full windows installation PLUS virtual server PLUS any virtual machines. If you are running virtualization in the core you take away alot of overheard from a host OS.

    As far as Virtual PC and Server being axed, isnt Veridian going to be free? If this is the case wouldnt Veridian just simply replace the virtual server product? So we have VPC left, Im gonna use my imagination and wonder if perhaps there could be some sort of update for Vista that would allow Veridian to run on Vista. After all isnt Vista and Longhorn practically same codebase?

    May 10, 2007 5:36 PM
  • May 10, 2007 5:56 PM
  • Wha? said:

    Dude, are you off the meds?  The quality of posts has really took a nosedive.    The core virtualization team is in India but the team is composed of MS employees.  So it's not outsourced.  It's just developed in a different location other than redmond.

    MS Team Foundation Server is mostly developed in Raleigh, NC.  Is that outsourced too?

    May 10, 2007 7:48 PM
  • MikeB said:

    So perhaps oOff-shored would be more accurate than outsourced.  

    Regardless, don't discount Windows Server Virtualization.  There is more than enough still in the product to keep it very competative.  The superior driver model, performance improvements and impoved memory management alone make it a very compelling and lower cost alternative to VMWARE.  

    Not everyone is willing to pay the premium for VMOTION.  

    Customers are still demanding an integrated virtualization solution from Microsoft.  I believe this gets them there.

    It's so easy to look at the loss of these high profile features and declare failure.  Nothing could be further from the truth.


    VMWARE customer looking to reduce cost and hopeful WSV TAP Customer

    May 10, 2007 8:51 PM
  • May 10, 2007 11:14 PM
  • Shawn Oster said:

    The only issue I have is with, "This "resources" thing is a load of crap. If Microsoft really saw virtualization as "the next frontier", they would bring all their weight to bear on the market..."

    I've worked on several projects over the years that just could not benifit from more resources.  Unless you've worked on a team developing a large project you probably can't understand this (heck, most managers don't) but you hit this saturation point where adding more developers actually slows you down.

    As a software developer yourself I'm sure you've hit those points where you were almost done except for one or two issues, and those issues can sometimes take days, even weeks to resolve.  If some programmer, even a damn bright one, walked in during that phase and asked what they could help with you probably couldn't say, "just track down this issue".  You'd have to bring them up to speed and share with them what you'd already tried.  They'd probably have a ton of questions about your framework or would start changing things that broke regression testing.  The problem is about a 100 times worse than that on a large project with many interlocking layers.

    Until you've had to play shepherd to flock of new devs thrown at a project because upper management wanted something sooner you can't understand the horror at hearing the words "just throw more resources at it."

    May 11, 2007 2:14 AM
  • Josh Bernard said:

    "...but supposedly the other editions will have that role as well. "

    By other editions are you saying that the consumer editions will include this or other server editions?

    May 11, 2007 8:57 AM
  • McLaws: you're doing fine. Your post count is fine, and the content is fine.

    H L: you're mistaking contrariness for reasoning.

    1. The timelines are _not_ on schedule for the published feature set.  While timeboxing works as a project management strategy, it's not an effective marketing strategy.  Heck, I can ship Windows 2010 today if I simply remove all of its features.

    2. VMWare's release of 6.0 Workstation shows they're still delivering value on the desktop, in addition to the work being done in ESX (and the still-free Server).

    3. You are lost in the fog of your own parallel universe.  I claim that my pet monkey will do all of these things (with 90% coverage) by 2008 as well.

    4. Of course this is true.  But it's also nothing new.   Software development companies (like Microsoft) have known this since the beginning of time.  To claim that this is the reason for Microsoft's slip is equivalent to explaining away poor battlefield performance by an army because they couldn't concentrate due to people shooting at them.

    Agreed that live migration is the biggest loss.  My VI3 environment makes possible supporting my current set of ~350 VMs (running approx 400 biz apps).

    May 11, 2007 10:32 AM
  • mcepat said:

    It is a shame, I recently went to the LongHorn roadshow and the track I took was showing all the upcoming features, live migration, hot resource add etc, it really would have set them apart with competetion, but now that I hear this not so sure since all the features I saw and drew applause is being cut?

    This is a very stupid turn if you ask me, the future of computing itself is Virtulization, not farms of physical boxes and not even blade servers, its big heavy duty boxes running 100's of virtual machines and from what I saw Microsoft has a really sweet management tool to take care of virtual server sprawl.

    but man they should do a about face like they did with the internet, the future is VM's and they should go 100% in this direction for the enterprise, where they already own the os and the server infrastructure software it seems the most likely direction you would think? Redmond? ae you listening?

    May 11, 2007 12:31 PM
  • H L said:


    1. Since last year... Microsoft always stated within 180 after the the RTM of Longhorn. As IT people, we expected more and made assumptions. If we are not happy with the timelines, we need to swallow the harsh reality pill that assumptions are the mother of all f**kups. As an enterprise customer and a Windows engineer for the past 6 years, I are VERY aware of the issues with Microsoft's inability to deliver products on time. I am also painfully aware of the reality behind their lack of product in the virtualization space. So if anyone was betting the farm on Microsoft, then you deserve what you get. If you invested properly in the VM platform, working with Microsoft to help them produce an enterprise ready product is part of the job. They still don't understand the enterprise as well as they should. Its our job as the customer to help them realize this and beat them over the head with our requirement. I am not sugar-coating the announcement... this is a massive problem for them and a deployment blocker for many customers...

    2. VMWare Workstation is a great product. But we are talking about enterprise-class products. Host-based virtualization (MS Virtual Server, VMWare Server) are not the strategic direction for server consolidation, hosted desktop, test/dev, dynamic datacenter. Stay on point with the thread...

    3. I will assume you know Virtual Center and ESX 3.0 very well. I will also assume you have worked with Microsoft on SCVMM from its inception as alpha. I will also assume you are on the Viridian TAP, I will also assume you have been performing density tests with the latest builds of Virtual Server. I will also assume that your thinking about VM technology is forward thinking ... >12 months. If all of this is true... and you still think that I am "lost in the fog of your own parallel universe" then so be it.. otherwise... hehe... you're just in the dark.

    4. Of course its known... yeah... its not anything new. I am not justifying anything on this... this is simply reality. We have to swallow the pill... get a hypervisor product on time or wait for all the features. This will help companies make strategic decisions. Not everyone is knocking on VMWare's door. If you are honest about the situation you know every well that many companies cannot justify the cost of VMWare ESX compared to their virtualization plans. Many are just getting their feet wet. However, the announcement does hurt Microsoft's entry into the enterprise... this is going to be the most problematic of the areas and an unfortunate development

    I would bet that you could virtualize and gain hige benefits even without VMotion. I know VI3 very well... the key to success for ANY VM deployment is planning and proper alignment of internal groups and expectations. VMotion is a cool technology that VMWare coined in the distributed space. But having it does not enable virtualization and does not solve world hunger.

    May 11, 2007 1:20 PM
  • May 22, 2007 5:50 PM
  • Sterghios said:


    June 6, 2007 4:14 PM
  • Alexander said:


    June 13, 2007 1:41 PM
  • Zaharias said:


    June 14, 2007 4:57 PM
  • Cosmo said:


    July 10, 2007 10:05 AM
  • Haralambos said:


    July 10, 2007 11:18 AM
  • Efstathios said:


    July 10, 2007 6:13 PM
  • Yannas said:


    July 10, 2007 11:39 PM