Ilya Baimetov's Server Virtualization Blog

When it comes to defining the role of virtualization, there are two camps: 1) Virtualization is a platform – represented by VMware; 2) Virtualization is a feature – represented by Parallels, Microsoft, Citrix, IBM, SUN, Novell, and Oracle.

VMware knows that only software companies that are platform vendors survive in the long run. VMware’s only asset is virtualization, so they have to position their own virtualization as replacement for someone else’s existing platforms. In this interview, Mendel Rosenblum says that hypervisor will not replace the OS, but in the next sentence he says that the OS as we know it, will cease to exist. Instead, it will be reduced to an application-specific library. VMware prays for it, while others are concerned – but this subject deserves a separate post. VMware goes against all established players, makes huge bet with a huge risk – when the base virtualization technology is free – and it will be free when Hyper-V and Xen, both free products, become more mature – VMware will have hard times charging the price premium they charge now. Their management suite only works with their hypervisor, while in the system management world being heterogeneous is the basic condition for success.

Second camp is saying that virtualization is optimization of existing platform. For Parallels, virtualization is a very important but only a component of the overall automation platform. For Microsoft, Citrix, IBM, SUN, Novell virtualization is very important, but it is still only an enhancement of their existing platforms and ecosystems. Parallels Virtuozzo and Parallels Server are a big part of Parallels automation/SaaS platform, but not really the main value. Hyper-V, Xen and others alike are very important capabilities, but again, not the main value of what the established players offer. When Simon said that virtualization should be a part of, not “the” infrastructure and all the disruption caused by VMware products is not necessary, an anonymous (why???) VMware employee went berserk.

I’m with the “feature” camp. In my opinion, virtualization layer is more or less similar to the OS kernel – it’s a foundation for everything else, but it’s not a deciding factor for choosing a platform.

Just as very few people would decide to buy a car based on how good the navigation system is, even fewer people would make their Linux vs. Windows vs. Mac decision based on how fast each OS creates a process or performs a context switch. Rather, they look at the value of entire platform and total cost of ownership.

VMware tries to convince otherwise. For example, take a look at the argument about importance of a patented memory enlargement method. The memory ballooning feature allegedly helps greatly improve the density of virtual machines and save on the hardware costs. However, as many opponents pointed out, it was a great debate about something not really important. Memory is cheap and hardware costs are a minor components of the TCO. Management cost is by far the biggest TCO component and VMware suffers from VM sprawl with or without memory ballooning. Parallels Virtuozzo, on the other hand, eliminates VM sprawl and delivers huge TCO savings. Not to mention that in situations where density really matters – light-weight web/LOB applications or VDI – Parallels Virtuozzo, with its true dynamic memory allocation, consistently delivers at least 3 times higher than ESX on VDI scenario and at least 5 times higher for light-weight web apps.

Anyway, what do you think – platform or feature? [Leave your answer here.]