The post entitled “Five Reasons Why VMware Virtualization is Better than Microsoft Hyper-V” on DABCC is already more than two weeks old. Perhaps I misunderstood some of the assertions in the article, but it seems to me that there are quite a few mistakes in it. I googled the topic to see if I had missed some essential news about Hyper-V; however, I wasn’t able to find anything that contradicts my view. So I decided to blog about it now because I am quite puzzled by the article’s arguments. Maybe you can tell me what I misunderstood.

I will reproduce the heading of DABCC article and sometimes a quote from Douglas Brown’s arguments, and add my comments afterwards.

1. Hyper-V is marketed as free but really isn’t.

Yeah it’s $28 bucks but that’s on top of a Windows 2008 License paid in full.

This is the first time I’ve heard that Hyper-V is marketed as free. At the moment, Hyper-V is just a feature of Windows Server 2008, like IIS7. Nobody ever said that IIS7 is free, so why should Hyper-V be free?

Anyway, the main mistake here is that Brown confuses two different products. According to this Microsoft press release, it is not Hyper-V, but Hyper-V Server that will cost $28. Hyper-V is a part of Windows Server 2008, and Hyper-V Server will be a hypervisor-based virtualization product which will be delivered by hardware vendors. As far as I know, it is not available yet.

I think what Brown wants to say is that ESX is cheaper than Hyper-V (not Hyper-V Server) if you add the licensing costs for Server 2008. However, the real price depends on many factors. It depends on the number of servers you want to virtualize and on the Windows server edition you run on the host system. For example, if you buy the Enterprise edition, you get 4 Windows Server licenses for the VMs and with Datacenter edition, you don’t pay anything for the Windows Sever guests. It also depends on the VM management product you will use. So I think there is no general answer to the question of which product is cheaper. It depends on your environment. If you intend to virtualize only Linux servers, you probably will pay less with ESX.

2. Hyper-V is not actually bare metal.

If Hyper-V isn’t bare metal, then ESX isn’t either. The only difference between ESX and Hyper-V here is that VMware uses a modified version of Red Hat Linux in the root partition and Microsoft uses Windows for this purpose. Of course, the advantage of Hyper-V is that it runs on every PC that supports hardware virtualization because there are more device drivers available for Windows. This is certainly an advantage of Hyper-V.

3. MS target market is low margin with little to gain because SMBs aren’t ready.

I doubt that Microsoft only targets SMBs. It is true that VMware ESX is technically advanced, but it is certainly wrong to put Hyper-V on the same level as VMware Server. VMware Server is comparable to Virtual Server. There is no doubt that Hyper-V offers better performance than VMware Server and also Virtual Server. Thanks to paravirtualization, it might even be faster than ESX in some environments.

However, my point is that when it comes to virtualization, Microsoft’s main product is not Hyper-V, but Virtual Machine Manager 2008. It was quite a smart move to support ESX in VMM 2008. Most virtualization experts agree that the hypervisor will only play a minor role in the upcoming battle between VMware and Microsoft. What counts are the management products. VMware is still leading here though. But we all know that Microsoft is not too bad at developing system management products.

4. Hyper-V ties the hypervisor back to the OS.

I already said it above. There is no real conceptual difference here between ESX and Hyper-V. If you worry about the overhead of a bloated Server 2008, you can run Hyper-V on Server Core. However, I suppose that Brown is referring to the fact that hardware vendors integrated the ESX hypervisor in their products, so he writes:

MS will have to release a true bare metal product to compete with VMware and they need to do it soon. I highly doubt they will though as Windows is integral to Hyper-V’s ability to do half of the things ESX does.

I wonder what reasons he has for these doubts. Didn’t Microsoft announce that they are already working on a “bare-metal” edition of Hyper-V in this press release?

5. Virtual Machine Density is a fraction of what is possible.

In this argument, he refers to the memory resource management features memory overcommitment feature of ESX. As far as I understand it, the main idea here is to virtually assign more memory to your VMs than is actually physically available. It is certainly a matter of fact that the amount of memory a server requires varies over time. Thus, with this technology you can run more VMs on a host than you can without it.

I must admit I have no experience with this feature, but I doubt that I would like to use it in a productive environment. The problem with any virtualization technology is that you run software in an environment for which it wasn’t developed. This can cause unforeseen problems. The farther you move away from a standard environment, the riskier it will be. And this technology sounds very freaky to me. I wouldn’t even configure virtual disks to grow dynamically in a productive environment. The main issue certainly is performance here. However, I also feel much more comfortable if VMs can rely on the resources that were assigned to them.

Anyway, memory resource management is only one feature of ESX that Hyper-V doesn’t have. There is no doubt that VMware is still ahead of Microsoft. The main point of my article is just that the five reasons given by Brown are not really the best ones. I do believe that there are good arguments for preferring ESX over Hyper-V.

I also do not want to rule out that some of my assumptions are incorrect. As I said before, one of the reasons why I wrote this post was because I am puzzled by the DABBC article. Please, let me know if you have any other information.