Comparing Microsoft and VMware Virtualisation


I was just reading Yankee Group's report - "Virtualization Price War: VMware's Little Big Horn?", and just had to share this with you (I think it speaks for itself):

clip_image002

If you haven't got access to the report (which we didn't pay for - before you ask), just ask & I'll get you a copy.

Dave.

Comments (31)
  1. So http://itcomparison.com/Virtualization/MShypervvsvi35/HyperVvsvmware35esx.htm is a great comparison of a production product (VMware) and a beta (Hyper-V).  So it’s kind of like comparing apples with oranges?

    I’d have to agree on the naming comment (I do not like Hyper-V)!

    Of course Hyper-V is targeted at test and development (at the moment) – it’s a beta!

    Of course System Centre VMM doesn’t manage Hyper-V (yet) – it’s still a beta!

    Of course we don’t support many OS’s on Hyper-V (yet) – it’s a beta!  Go look at the supported guests for our production offering (Virtual Server).

    Performance is an interesting one – what is performance?  How do you measure it?  Is the number of guests the real measure?  We already support more guests per host on Virtual Server.  Hyper-V is still in beta.  Hyper-V’s hypervisor runs at ring -1 on the processors, which actually means the guests are all running at ring 0 (even closer to the metal than VMware)!

    Live Migration vs. Quick Migration is always an interesting one.  How many customers "really" need close to zero downtime?  I don’t come across many (most just need the services to be up and to have the ability to move that service).  Quick Migration is quick (a few seconds).  How much are those few seconds worth?

    Booting from virtual SCSI disk – we already do that with Virtual Server.  Hyper-V’s virtual IDE and virtual SCSI are just synthetic device drivers – they are equally performant (i.e. this is irrelevant).  What’s  more important is the physical storage that we support (we support a lot more than VMware).

    Backup Performance – Hyper-V is still in beta!  Taking live snapshots of running machines is possible on both Virtual Server and Hyper-V in the box.  Both System Centre Data Protection Manager and Virtual Machine Manager can initiate the snapshots too.

    Special Hardware Requirement?   Pretty much every server that’s shipped in the last couple of years has been x64 – which means Hyper-V will run on pretty much every "modern" server.  VMware can’t say that – there is a very limited sub-set of hardware to choose from (because of the device drivers being in the hypervisor).  We already run (and will do with Hyper-V) on any server with the “designed for Windows” logo (i.e. any server).

    If you want to run a 32-bit host operating system, then use Virtual Server.

    Thanks for your comments guys – it’s good to talk!

    Dave.

  2. Martin,

    Sorry I didn’t reply to this – I’ve turned off the notifications of new comments on my blog (am getting a lot of spam comments).

    Sounds like a great idea.  I haven’t got two sets of any hardware (just my demo laptop).

    How about I come down to you – you work for VMware in Cork (right)?  I imagine you’ll have some kit lying around (and will be OK with breaking the EULA)?

    Drop me an email & we’ll arrange something.

    Dave

  3. Dave, you are probably making a lot of great comments here.

    Please don’t "shoot the messenger" – the report was done by the Yankee Group (not me).

    I posted the price comparison chart to get people thinking – it certainly wasn’t meant to be taken as the gospel (we all know that there are many ways to buy stuff – and to get free stuff).  We did recently change the price of System Center – you’ll probably find that it’s just our website that’s not up to date.

    My personal opinion?  I recon Virtualisation is only just starting (less than 10% of servers are virtualised) – there are some very "interesting" things happening (the Dynamic Datacenter, etc).   Who knows what will happen?  We’ll probably all end up running a modern equivalent of a mainframe and be using very intelligent "dumb terminals"?  I’m just happy to be working in IT during this very exciting time!

    Dave.

  4. I can run 7 VMs on my laptop!!  What are you doing Tim?

    Has anyone ever read the VMware End User License Agreement?  They won’t let you publisise your performance benchmark tests without their approval:

    "You may use the Software to conduct internal performance testing and benchmarking studies, the results of which you (and not unauthorized third parties) may publish or publicly disseminate; provided that VMware has reviewed and approved of the methodology, assumptions and other parameters of the study. Please contact VMware to request such review."

    I’d love someone to do the comparison tests and go public with the results.  It’s interesting that no one has done so yet!

    I still think that "most people" just need the ability to move a VM without losing client connections (i.e. a couple of seconds is fine).

    Keep the comments coming guys – this is interesting..

    Dave.

  5. Interesting comment from "Dave is a Jack Ass" – Apparently everything I have written about VMware is incorrect?  Interesting because I haven’t written anything about VMware – just posted a chart from Yankee group, an excerpt from VMware’s EULA and asked if people really ‘need’ less than a second to fail over a VM (which they don’t).  Everything else I have written is about Microsoft products.

    Perhaps "Dave is a Jack Ass" is referring to himself?

    Dave

  6. Anonymous says:

    Unfortunately VMware was not consulted on this report, and the Yankee Group report contains many inaccuracies. We’ve released some corrections here:

    http://blogs.vmware.com/virtualreality/2008/02/those-darn-de-1.html

    A short excerpt:

    Exhibit 2 indiscriminately mixes VMware’s bundle and a la carte pricing, making it very difficult for readers to get an actual apples-to-apples cost comparison. If the table is meant to be a cumulative account of Microsoft and VMware pricing (i.e. add up all the rows), then there is significant double-counting in the VMware column. If the rows in the table are meant to be interpreted individually, then there are gross inconsistencies between how VMware and Microsoft’s costs are shown. Most casual readers will assume the table represents cumulative cost for both companies since they see the word “Included” throughout Microsoft’s column.

  7. Hi dave I would love to see the report from yankee group. We are in the middle of choosing vmware or ms virtualization, and that would help.

    bill.bacoyiannis@mckesson.ca

    Thanks

  8. Jason Boche says:

    If it wasn’t for VMware’s innovation, I seriously doubt Microsoft would have gotten this far with their virtualization endeavors.  Recall Microsoft’s first virtualization product wasn’t anything that they had even developed – they bought it from Connectix and slapped the MS logo on it.

    Microsoft marketing claims dominance while standing on the shoulders of the pioneers responsible for setting the bar where it’s at today.  Microsoft’s virtualization technology isn’t up to standards yet.  Case in point, when Windows Server 2008 releases in the future, their technology STILL wont’ be up to standards because MICROSOFT HAS NO LIVE MIGRATION FEATURE.  I can’t put into words the flexibility that VMware VMotion affords production environments.  Invest in Microsoft virtualization and I guarantee at some point you will suffer virtual machine outages due to a scheduled or unscheduled Microsoft host outage.  The boss will want an explanation for this.

    But I get your point.  On paper, Microsoft virtualization is cheaper and depending on the company or environment, I’ll be the first to admit it may be more cost effective.  In their traditional style, Microsoft has been attacking VMware with low price points on their products.  Microsoft can afford it and they are buying their marketshare with their deep pockets rather than earning it with their technology and innovation.

    Jas

  9. Tim Kelly says:

    I disagree. VM has set no technical bar, and indeed has bought or licensed many of their so called innovations. But in true EMC/IBM style, VM has priced themselves right out of the market buy ensuring the most basic virtualization functions become another SKU. VMware was first in many areas, but the Hyper-V days are here, and now customers can have a choice, just as in Oracle v SQL days.

  10. kimono says:

    Haha, speaks for itself does it? That crummy table has me thinking three things:

    (1) MS are just using their usual tactics to get their shitty version 1 product out onto the market.

    (2) Where did those VMWARE prices come from?

    The simple fact that "Live Migration" is tagged N/A on Microsoft’s column is just enough for serious enterprises to fully discount it as an enterprise ready system. VMWARE raised the bar here and set the precedent. This feature in VMWARE is worth those larger sums of money listed in the VMWARE column.

    “Capacity planning” is a freely available feature to Virtual Center 2.5, this is a newly released feature and offers users of VC subset of the "mother of all" features available in the VMWARE corporate services hosted and run service. Also capacity planning as a generic feature is in both products, but it would be worth drilling into both products just to see what you get “out of the box”

    "Library" – Virtual Center templates are freely available.

    "Self Service" is not a feature of Virtual Center but has been the realm of many third parties. (VMWARE acquired Dunes to address this, and pricing remains to be seen).

    Dynamic Resource Management – Microsoft’s model is completely different to VMWAREs for the same reason of VM portability. There is no comparison, we can shift our VMs to another host if we want to. Microsoft rely on their own clustering at the guest OS level.

    Storage Transportability – VMWARE have introduced truly portable disk files now with Storage VMOTION. You don’t pay an extra cent for that new feature. That doubles that theoretical cost of VMOTION. Microsoft = cold migration. Shut the VM down and shift it.

    The cost of ESXs base hypervisor is also not easily compared to Hyper-V because you are paying for a completely different type of hypervisor. You can also buy it in many different flavours such as 3i, Foundation, Standard, Enterprise. Depending on how big you are

    So a much better argument would be found debating the merits of VMWAREs overall virtualization model over Microsoft’s. VMWARE have aimed for, and achieved truly portable virtual machines that can easily be moved. They have a far greater range of supported guest operating systems, tighter integration with SANs, a much richer availability of third party (and soon to be native) applications for Disaster recovery and High Availability. See some of the recent webcasts from VMWARE.COM. For many organizations things like VMOTION has become an intrinsic part of their operations, and the simple portability is also crucial to the DR plan.

    Bottom line, the question still remains which is the superior hypervisor. Which has a greater “richer” and more “relevant” level of functionality. The data center has changed and customers expect certain things. Nobody has probably had enough field time with HyperV to really give a valid argument. I sure would like to compare, but I’ve got a strong feeling which one of the two is superior. The writing is already on the wall, with major hardware vendors embracing hardware based hypervisors, enabling closer integration with hardware and HW management solutions. Hyper-V has missed that boat. (There’s one reason for offering it at a bargain discount clearance price.)

    Licensing costs for both Virtual Infrastructure, and Guest Operating systems are also trivialized when you consider that enterprise customers are also highly likely to have an ELA for these products, yes both VMWARE and Microsoft.

    It’s true that because VMWARE have been the main player for a long time, they have been able to dictate cost. So they will have plenty of evolution to do now they’re up against Goliath. But, it’s all well and good to say how Microsoft looks on paper. At least customers will agree that VMWARE have earn’t the right to charge that premium.   (as an enterprise we couldn’t give a toss if VMWARE didn’t drop prices. It’s an essential piece of infrastructure we want to pay for).

    The bottom line, speaking subjectively… I couldn’t think of anything worse to run Windows on than,… Windows.

    FYI There’s a much better comparison of the two products here:

    http://itcomparison.com/Virtualization/MShypervvsvi35/HyperVvsvmware35esx.htm

    kim

  11. Chris says:

    Speaks for itself? Hardly! As other posters have pointed out, this chart doesn’t even scratch the surface of the differences….

  12. Paul Browne - FirstPartners blog says:

    Competition is always good, for both VMWare and Microsoft.

    I would take VirtualPC more seriously if the Linux support hadn’t been removed when Connectix was bought over.

    Even now it’s unclear if non Windows OS’s can be hosted in the latest version on not.

    Any pointers Dave? Will keep an eye out for you at the IWTC.

  13. Dave Mishchenko says:

    Hi Dave,  I think they could have done a much better price of comparing the actual costs.  I use both Microsoft and VMware virtualization products, but to be clear I have a bias towards VMware’s products.  I’m sure other’s will comment on the technical merits of the products but I’d like to just focus on the costs

    1)  The charts states that Guest Licensing is included, due to the purchase of Windows Enterprise Edition (EE) license.  That’s a correct statement, but not really that practical in the real world.  How many customers are going into virtualization to just run 4 virtual machines, which the EE includes, on a single host.  Rather for low resources virtual machines a customer is looking to get 4 to 8 VMs per CPU core.  It would have been a more fair comparison to use DataCenter licensing, in which case either product requires Windows licensing of $3000 per processor so in a 4 CPU server the cost would be $12K per host for both products

    2)  The live migration / HA / vSMP  / Dynamic Resource Management pricing is incorrect.  Here’s a good source for pricing on VMware products – http://www.vmware.com/products/vi/buy.html.  With ESX Enterprise all those options are included in the price.  You do require the VirtualCenter Management Server but that is one time cost so if you have 10 servers in your host farm, then it is only $500 per server.  VMware also offers some bundle pricing with reduces the initial cost of VirtualCenter for people managing 3 server or less.

    3)  Management pricing – with VMware you have a one time cost of $5000 and you can then manage any version of ESX which runs in price from $500 (ESX 3i) to $5750 (ESX Enterprise).  Microsoft’s web site lists the price for the Systems Center Server Management Suite at $1290 (http://www.microsoft.com/systemcenter/svrmgmtsuites/howtobuy/default.mspx) and I would take it this would be per server (i.e. a CAL) and you would still have to license the server running SCCM, DPM, OM, etc.

    4)  Library – templates are included with VirtualCenter plus you have a large library of pre-configure virtual appliances available on the VMware and other websites.  Lab Manager as a product has no comparison to a server running Windows 2003/2008 with VirtualServer.  Note sure how they came to that conclusion, but you can be glad that you didn’t pay for that research.  

    Overall if you want to provide a useful cost comparison,  you should provide the overall cost that either solution would require.  Start with 5 servers (2 x quad core CPU, 32 GB RAM),  FC or iSCSI SAN, all the VMware / MS licensing (including  the management pieces) and then you’ll have a better idea what your real world costs will be.  Besides the gross inaccuracies in license costs provided here,  this chart doesn’t provide a realistic cost comparison.  

    Further,  if you’ve been around the SQL performance studies like this http://www.tpc.org,  you’ll know that it’s just not about how much, but how fast for how much.  The same applies here.  How many servers (say 2 quad core CPU with 32 GB) running MS Virtual Server R2 would you need to equal the performance of 5 such servers running ESX?  If you do that comparison, you may find that the premium you pay for ESX Enterprise is recovered.  It will be interesting to see the results of such testing with Hyper-V and ESX 3i,  but with Virtual Server 2005 you’re not going to match the performance of ESX.  To quote a MS case study of the largest Virtual Server farm – 1400 virtual machines on 375 hosts (http://www.microsoft.com/casestudies/casestudy.aspx?casestudyid=4000000664)  gives a ratio of just over 4 VMs per host.  The ratio drops when you consider that for production they run 1000 VMs on 275 hosts.  In fairness, the study does not provide sufficient information about the nature of the VMs being run, but a 4 to 1 ratio isn’t much to brag about.

    An aside,  I think MS would do us all a service by providing more info on Hyper-V on it’s website,  Half the stuff I find today is still related to the beta or RC and what applies to the released product is lacking in depth. Perhaps you could provide some good technical links for that.

  14. AWo says:

    Linux virtualization is provided by Xen in Win2008, so it’s becoming a pretty variegated thing…

    AWo

  15. v01d says:

    Are you kidding me?

    Daven

    "Live Migration vs. Quick Migration is always an interesting one.  How many customers "really" need close to zero downtime?  I don’t come across many (most just need the services to be up and to have the ability to move that service).  Quick Migration is quick (a few seconds).  How much are those few seconds worth?"

    This is a joke right?

  16. kimono says:

    Plenty of customers need close to zero downtime. If Microsoft can’t see that, they’ve really left their blinkers on.

    Here’s another good wrap :http://blog.scottlowe.org/2007/07/23/live-migration-vs-quick-migration/

  17. ayoung says:

    "it speaks for itself"

    It certainly does. It tells me that the Yankee Group have quite clearly failed to do their homework. Anyone who has studied these competing products (1 of which doesn’t even exist yet…) for more than 5 minutes would quite quickly identify that the above table is blatantly misleading at best.

    The fact that the report is being highlighted by an employee of Microsoft (who should know better) is disappointing in the extreme, and provides some credibility to VMware’s claim that MS is deliberately confusing the marketplace with FUD.

    And as for your comment regarding live migration not being a requirement for enterprises, how else can you perform intelligent load balancing across physical hosts? Or do Microsoft think it fine to kill a VM for a few seconds (probably during business hours)each time  load needs to be redistributed?

  18. William Bishop says:

    I want some of whatever drugs the idiot that wrote this drivel is on. Must be good stuff. Can they be sued for incompetence?

  19. Azn2kew says:

    Frankly speaking, there could be better virtualization product than Microsoft trying to bring on now.  BUT, I can tell you for sure VMware is on top of the market for enterprise virtualization deployed out there with large corporations.   As far we know, in 2010, they will be 85% enterprise business deploying virtualization and VMware took almost majority of the market shares. I’m curious, the editor has ever use M$ and VMware VI3 product in real world?  Has he ever try to consult with ESX expert out there and know in depth with VMware licensing piece before he send this comparison chart? ESX expert speak for yourself.

  20. Darth Vader says:

    If you compare the free VMware product with HyperV, you’ll see that you’ll have a more mature product and one that works. How do you even know HyperV will work? Look at Vista SP1!

  21. kimono says:

    I was just catching up on all the responses on this blog over the weekend, and got a page not found for 15 minutes there. I presume a live "quick" migration was in process? Was the host down for maintenance?

    That just emphasizes one of the main points!

  22. Jason Boche says:

    Add this information from a recent Microsoft blog to the Yankee Group report:

    "Operating system installation on Hyper-V is slower than on Virtual Server / Virtual PC."

    "I do not expect this to change much for the first release of Hyper-V."

    Source:

    http://blogs.msdn.com/virtual_pc_guy/archive/2008/02/26/hyper-v-and-slow-guest-os-installation.aspx

  23. wildcattdw says:

    In my mind, the biggest thing that the comparison missed is density.

    Dave, to respond to your question about performance ("what is performance?  How do you measure it?") Seriously? There are probably seven ways from Sunday that you could accurately compare performance between the two products. I like to compare guest density. I have several servers that are around 18 months old (quad socket single core) hosts that are running 20 – 25 guests, and some ES7000/one hosts (quad socket dual core) hosts running 40 – 50 guests. That equates to 6.25 guests per core. In my testing of 2008 Server w/ Hyper-V (and don’t give me that "beta" status crap. If it’s good enough to release for testing, it’s good enough to use for very general benchmarking) on a dual socket dual core/24 GB host, I have only been able to run 8-10 machines; and of those I can only run Server 2003 SP2 or Server 2008.

    Any VMotion/Live Migration is HUGE. To downplay that is almost a crime.

    Tim

  24. Dave is a Jack Ass says:

    Dave,

    where did you get your information. most everything on the VMware side is incorrect

  25. kimono says:

    2nd attempt at posting this, the blog timed out again.

    This is well worth a look.

    http://vmwaretv.blip.tv/#745090

    Quick Migration is nothing but an orchestrated fumble, in no way is it seamless, or even close to true portability

    We’re not just advocating VMWARE VMOTION Dave, there’s a massive community based around it. Unfortunately all MS have is the marketing clout to push their inferior solutin.

    KIM

  26. Tim says:

    Kim…  in your video…   how much RAM was your VM and what backend fiber were you running on?

  27. 1800f says:

    2t2MB9 gdfgdfg dfgfdgfd gfdgfgfd

  28. flow1800 says:

    ZpRvNA fgadioghrg nuiowrhg489gsd gnsdjig djshgjdksh gjkdshgjksd gsdgds

  29. flow1800 says:

    U8WmCG htdioghrdeioh uerhuy h3oh iohgdfiogho isdfhio sdjhdf

  30. Martin Doyle says:

    Dave

    Don’t know if you remember me, I met you at one of your Microsoft Virtualisation Seminars in Cork a couple of years ago when you were presenting Veridian (now Hyper-V).

    First of all, congratulations on the full release of your Hypervisor, hope that all goes well for you and Microsoft.

    I have been reading your Blog posts over the last couple of years and following the Virtualisation blogs and comments with some interest. When we met, you were talking about VMware’s EULA where it says "You may use the Software to conduct internal performance testing and benchmarking studies, the results of which you (and not unauthorized third parties) may publish or publicly disseminate; provided that VMware has reviewed and approved of the methodology, assumptions and other parameters of the study” and I also see that you have blogged about this since.

    As I said to you at the time, the reason for this is because quite simply that people who do not understand Virtualisation and its architecture and have published “benchmark” testing comparing VMware’s Virtualisation with different types of architectures, i.e. comparing “Apples” with, well, “Potatoes” or "Carrots". You made a comment on that blog that “I’d love someone to do the comparison tests and go public with the results” so now that Microsoft has released a Hypervisor and as such you are now in a position to compare “Apples” with “Apples” I am offering to work with you to do just that.

    I am based in Cork but am in Dublin at the moment and will be for the rest of this week, I can find some time to call into you in Sandyford and do a quick comparison between ESXi and Hyper-V and publish the results live online.

    Let me know what day you are free where we can put ESXi and Hyper-V on a couple of identical physical infrastructures side by side with a couple of servers, shared storage and networking, and run through some benchmarks. While we are at it we can also have a look at the differences between VMotion and Quick Migration and compare the actual costs for each solution, putting to bed the Yankee Group’s report you bolgged about.

    Looking forward to meeting up with you again.

    Regards

    Martin

  31. Martin Doyle says:

    Dave

    Thanks, just saw the reply, working on a big project in dublin at the moment, will email you next week.

    Martin

Comments are closed.

Skip to main content