Hello everyone. This week, I want to cover a topic that has been discussed on numerous blogs, articles, and even tweets over the last several weeks. The topic is the cost comparison between Microsoft and VMware virtualization solutions. As many of you know, we often cite the fact that for server virtualization, VMware's solution is 6x the cost of the comparable Microsoft solution.
The Microsoft Comparison
So how do we get to those numbers? The calculations and how we get to those numbers is right on our cost comparison webpage. On that page, we show a chart of the costs and features surrounding the comparison. If fact, you can try the calculations yourself using the Cost Comparison Calculator. To make it clear, this comparison uses the following parameters:
With those parameters, the results are pretty clear and the cost differential is very significant. Some people point out that the vendors will discount their software. That is true but the levels of discount differ amongst customers. But just a quick check shows that even if VMware gave a 70% discount on their list price, the VMware software still costs almost twice as much as the Microsoft solution at retail price.
This cost comparison also allows for one of the closest, apples to apples comparisons. The cost comparison is based on the key differences that apply for most implementations, being the actual software and licenses that a company would have to buy to implement each individual solution. It should be noted that all the features listed in the chart are included in the cost listed. This is important because most VMware cost comparisons use the base cost figures but most VMware feature comparisons use features (such as Site Recovery, Lab Manager, AppSpeed) that cost thousands of dollars per host, the cost of which is never added to their comparisons.
There are other potential parameters, as we’ll discuss later, but the costs compared here will always be different between the Microsoft and VMware solutions, something that can’t be said of other parameters. If you want to do your own comparisons, changing server numbers or different software editions from each vendor, you can use the Cost Comparison Calculator that is available off the Microsoft Virtualization website.
One question we get is why we didn’t include the cost of Windows. The reason for this was based on feedback from our customers and partners. It was pointed out to us that some customers already own Windows, so that cost isn’t involved. Even for those who don’t own Windows, they would buy the Windows licenses for both solutions. Finally, with our R2 releases, we have made Microsoft Hyper-V Server 2008 R2, our free, standalone hypervisor, available with true feature parity with the Windows Server 2008 R2 based Hyper-V. Thus, customers can choose either Microsoft Hypervisor solution, have the same features and benefits, and in the end, the same net zero cost, allowing the removal of the OS cost from the calculations.
Sharp eyed followers of this discussion will also note the cost comparison went from 1/3rd last year to 1/6th now? How did that happen? Two reasons: We removed the cost of the operating system (which you can add back in via the cost calculator) and VMware introduced their Enterprise Plus SKU, which increased the cost of VMware for most Enterprise customers (since many of the newer features of vSphere are only found in Enterprise Plus and VMware is retiring the older, Enterprise SKU).
Not surprisingly, VMware doesn’t agree with our assessment. :) VMware challenges our statements but don’t actually challenge the cost comparison, as the numbers really speak for themselves. VMware presents an alternative cost comparison method, which they label as Cost Per Application. There are two very simple reasons why VMware went with this calculation. First, by including additional costs such as Real Estate and Power, they attempt to cover the true cost differential in terms of the software purchased. The second is that VMware then uses a white paper study to justify that VMware can run more VMs than Microsoft, claiming up to a two to one ratio.
Let’s take a look at the first concept, the use of cost per application. Like any other unit of comparison, a cost per virtual machine comparison is only as useful as the parameters it is based on. In VMware’s case, there are several issues of note that you should consider.
The most glaring parameter missing is that capability of the Microsoft System Center solution and the need to factor equivalent costs for the VMware side. While VMware argues that their management solution is better than Microsoft (which we don’t agree with), VMware completely ignores the solutions that System Center brings to the application and service level. This includes complete Patch Management, Software Distribution, Compliance, Inventory, Monitoring, and Backup and Recovery, for BOTH physical and virtual systems, from hardware to software, all the way to the service level. To get the equivalent software solution for VMware, one would have to buy a significant amount of third-party software, most of which are licensed per VM, along with that associated infrastructure. In fact, one of the most cost effective management solutions for VMware is the System Center solution, resulting in adding the System Center license cost and infrastructure to the VMware calculation.
The second concept is the use of memory overcommit, a feature that allows the running of VMs with more memory than is physically available. The concept is essential to the calculator, as without it, the VMware solution just isn’t cheaper than the Microsoft solution. Matt McSpirit, a Microsoft Technology Specialist, covers this pretty completely in his own response.
The increased consolidation ratios shown in the calculator is based on a single, VMware-sponsored study. For an in-depth look at memory overcommit and the study behind the calculations, please read the Virtualization Reality white paper. This white paper reviews the study itself, what the overcommit is, and what are the potential issues with it’s use.
Finally, the calculator doesn’t factor in the most important variable for a Cost Per Application calculation, which is the the application. Without factoring in the application to be run, you can’t actually calculate the hardware and software cost or the consolidation ratios. How much memory the VMs take, how they run, what additional software you might need, is all dependent on the applications that actually run in the VM. If you factor in applications into the calculation, you get a much more accurate calculation but the results cannot be applied, implementation wide, to all VMs, which is what VMware is trying to do.
If the discussion on the merits of VMware’s calculator seems complicated, it is because it is complicated. More importantly, while the calculator might be useful to estimate potential costs for virtual machines, what it is not good for is comparing differing solutions. In the end, the final cost of implementation varies by company, technology, and applications. Regardless of the various parameters, I’m confident that you find that the Microsoft solution will save money over an equivalent VMware solution.
And linux and bsd will by same logic save money by being 100x cheaper than microsoft. But then microsoft would calculate things differently claiming "total cost of ownership".
Personally I do use *free* vmware player, makes it safer/easier to run older software, and can have exact same setup on multiple computers if needed.
Cost comparisons are really only valid if the given comparison results in a similar comparison. That's the point I was trying to make in my post. There is a difference between cost comparisons and cost calculations. Cost comparisons are harder to normalize because applications and usage impacts the actual costs.
There are a lot of free virtualization options out there, from Microsoft, VMware, and others. They have specific advantages and disadvantages. VMware player is a good solution for what it does but if you are an Enterprise customer performing high-end Server Virtualization, the VMware solution (vSphere) is very expensive.
My company has used both virtualization solution (Microsoft and VMware) and in the end we found out that VMware suits our requirements best and Microsoft Hyper-V lacks the stability(typical featuer which comes free of cost with all microsoft products).
I think makers of Hyper-V were not able to provide basic Microsoft feature (CTRL + ALT + DEL).
What say readers?
While no product is perfect and rarely suites all users, I think that Hyper-V has show to be a excellent Hypervisor. As I noted my other posts, it's not an all or nothing choice. End-users can choose to use Hyper-V along side of VMware, all while managing it using System Center.
I would encourage all who want to research Hyper-V to see the case studies we have at http://www.microsoft.com/virtualization.
If there are specific issues you have run into, please let me know and I would be happy to try and help. As for Ctrl-Alt-Del, that is not available by default when accessing the VM because that key combination could affect the host, not the VM. This is true for almost all virtualization software, including VMware. You just need to use the alternative key combination, which for Hyper-V, is Ctrl-Alt-End.
Ok, so VMWare costs significantly more than MS's solution, for LICENSING. So what?
Would you recommend using ONLY Terminal Server over Citrix in an Enterprise environment because TS is so much cheaper?
If you answer yes to that, remind me to NEVER hire you, because you refuse to acknowledge the real key is TCO.
There are simply vast differences between SOHO systems (MS designs) and Enterprise.
Mind, I'm not criticising MS here...over a decade ago they stated their target market was Small-To-Medium businesses (which I think was a VERY smart move). To that end, their products are designed mostly for that market, with extendibility added for the enterprise.
I couldn't imagine running our company on HyperV. It's memory utilization alone would cost us WAY more than the licensing differential for the same performance. There are limits to how much memory a single blade can support, and once that memory wall is hit, our ONLY expansion is a new blade ($10k). ANd if our chassis is full, that next blade will cost us a new chassis too ($30k-$60k).
So, Hyper-V is still 6x cheaper?
As a former Citrix Winframe Admin, I know what you are trying to say. That being said, the focus of this specific article is on how cost comparisons can vary. TCO is a key factor but the issue is that TCO is based specifically on each implementation, which I'm sure you'll agree on. The problem here is that TCO is done at a detailed level and is not something that be can stated or declared at a marketing level. That's exactly what the entire second half of the blog is about.
Another factor is the assumption that Hyper-V isn't able to meet the needs of Enterprises. While it may not meet the needs of all Enterprises, there are numerous examples and case studies which show the capability and effectiveness of Hyper-V for the Enterprise. I've written several other posts, on this blog and others, about the fatures and capability of Hyper-V in comparison to other solutions.
As for the memory utilization, I would suggest you look at the whitepaper I referenced in my blogs post and also Matt McSprit's post. Again, final levels of the use of memory overcommit relies on the actual Guest OS and applications that run within them.
In the end, the final cost TCO is dependent on the customer and implementation. We note the inital capital cost because it is the closest apples to apples cost comparison that does not have a dependencey on the workload.
The issue I wanted to raise is for people to TRY Hyper-V and do the analysis for themselves. Only then will a comparison at a TCO level be applicable.
You have to stop with the 2x, 3x, 6x the cost lies, you know that your cost comparisons are not relevant in the real world, rather they are some marketing numbers made up to make you look good. VMware has an extremely thorough calculator on their web site - http://www.vmware.com/technology/whyvmware/calculator/ that shows they are cost equivalent when running just one more VM per host with their top of the line product and cost equivalent on more comparable products at the same consolidation ratios, where is the 2x, 3x, 6x in that comparison??? Man up and compete with VMware with facts rather than some arbitrarily made up numbers. Additionally SCVMM does a horrible job of managing a vSphere environments in our testing, so your claims of being able to manage both Hyper-v and vSphere are simply not true.
I completely agree with the blog. Each environment is different. From an apples to apples perspective, VMWare is far more expensive than Hyper-V up front. Especially once you start adding the things you need for management, HA/DR, etc.. I've implemented both VMware and Hyper-V solutions and think Hyper-V is fully capable of Enterprise implementation, provided you get the SCCVMM, which works great for migrating from VMWare to HV. However, I personally ran into the issue of losing a couple of guests (due to performance) when we switched from VMWare 3x to Hyper-V 2008(pre-r2), thus the money we saved was lost by adding another host. Like I said each environment is different and you have to do your own calculation to come up with a percieved value of the underlying solution. Also, having been a Citrix / TS admin for the last decade, I suggest you take a closer look at TS/RDC that comes with 2008/r2. There is really no reason to implement Citrix anymore for a published application / remote desktop session. I'd give Citrix credit where it's due but my headaches went away once we migrated to a TS 2008 environment, particularly printing and client deployment. I'd say it's the best thing MS has done in a Looooong time...
I agree with the blog. Citrix Xenserver need to considered in this scenario as an alternate Virtualization solution for both Hyper-V and Vmware.
Thanks blogger and posters for your discussions here.