I believe that KVM will be a significant player in the virtualization market. Here's why:

Birth of Virtualization

Virtualization has been around for a long time. We first saw virtualization technology from IBM in the 70s as part of the System/360 project called the IBM VM. One key feature of the System/360 was that it allowed customers to consolidate a few System/360 minicomputers by upgrading to a single System/360 mainframe, assuming the mainframe could provide the same features. It was able to achieve this by using the concept of a self-virtualizing instruction set. Consolidation is still one of the most attractive features of virtualization technology today. Back then, computers were limited to business customers for the most part and it would take the explosion of the Personal Desktop before virtualization could become mainstream.

Entrance to commercial market

Fast forward 30 years after the rise of the PC and enter vmware. My first experience with virtualization began with vmware version 2 around 2000 but vmware had already been around for a few years by then although still relatively unknown. It was the year 2000 when I overheard someone speaking about a technology to “run multiple operating systems within an operating system” and thought I needed to have a look at this. By this time, vmware was starting to pick up momentum in the commercial market for desktop use. What I found most useful was using vmware as a “lab” to test software and server configuration and even testing client server interactions right on my home computer or laptop before deploying in a production environment. I was hooked. For the next few years, vmware held the spotlight and had no real competition. They shortly after introduced their enterprise GSX product and later their “bare metal” hypervisor ESX product. ESX was designed for the data center and ruled the datacenter kingdom. It was the only ‘bare metal’ hypervisor around and was a lone player in a potentially huge untapped market.

Chip Makers Influence on Virtualization

When you think about it, it is really the chip makers, particularly Intel and AMD, that have driven the growth of virtualization in the commercial market and not really Vmware. Here’s what I mean. Virtualization ( which is really partitioning ) is unrealized if you don’t have available hardware resources to partition. It doesn’t make sense to partition a 386 PC when it really only has enough cycles to run one operating system comfortably. Intel and AMD have been increasing the power of the cpu for PC/Servers exponentially to the point where there was enough spare cpu cycles to start partitioning into virtual machines. By this time, vmware was already the leader and was perfectly positioned in the market. Their timing was perfect.

The chip makers, while they helped propel vmware to the leading hypervisor through increasing cpu spare cycles, are also dealing vmware a lethal blow – Commodity Virtualization. Vmware has played an important role in commercial virtualization. They have provided stable, high performance virtualization using software only and has done a superb job at this. However, as any other industry, commoditizing certain parts of an industry is the best way for the industry to grow. The chip makers have started this process by providing hardware virtualization. They have already partitioned the instruction set and some IO in hardware. Soon will be memory partitioning. By further partitioning more hardware resources, software hypervisors will become increasingly simpler to write (relatively speaking) and will become the next commodity. This is where vmware will suffer.

OpenSource Virtualization

While hardware virtualization is becoming a commodity, we are beginning to see a proliferation of hypervisors in the market which is a natural outcome. The first promise of an opensource hypervisor was Xen. However, Xen’s initial focus was on paravirtualization which will become less competitive with hardware virtualization and paravirtual device drivers. Xen now supports hardware virtualization , however with its recent acquisition by Citrix it looks less attractive to the opensource community. Enter the latecomer KVM opensource hypervisor which launched in 2006. Although late to the gates , KVM has many advantages which I think will eventually position KVM as a major player in the virtualization market.

Why KVM will be King

After that longwinded historical introduction, I can now list my reasons for why I think KVM will be king in context.

1. KVM is already a commodity. Part of the reason for the dotcom bust was that many companies were selling technologies and not services. Technologies eventually become commodities. An example is the telephone. The basic telephone technology is dirt cheap. What you pay for are the services such as call waiting, three way calling, voicemail etc. Similarly, in the virtualization industry, the chip makers are commoditizing the technology. What will eventually sell are the services and application on top of the hypervisor. KVM is already part of the linux kernel base which means it is a proper commodity as a software hypervisor. This means it will see more installs by default as linux has a large install within data centers and wherever there’s linux, there’s kvm. Even cisco is using kvm in its latest carrier class routers. See here.
2. It is simpler than the leading hypervisor. Vmware’s flagship product ESX is a microkernel operating system onto itself. It has to deal with scheduling, resource management, device drivers and other issues that are operating system related. In addition to this, ESX cannot support all hardware because it is bogged down with developing device drivers and even installs it’s own proprietary filesystem. KVM relies on Linux for these OS services and device drivers and therefore will support any hardware that linux device drivers support - which is significant. No special filesystem is required which means simpler management as far as backups and migration. KVM can then focus only on hypervisor features. This means that the rate of development will be much faster.
3. KVM has greater mindshare. Because KVM is opensource and part of the linux code base, it will have more eyes looking at it and contributing to it. Apart from faster development, it will be under more scrutiny and therefore higher quality; a benefit of the opensource development model.
4. Industry support. KVM is supported by major industry players such as IBM, cisco, intel, AMD, Redhat, novell amongst others. The support of such companies to an opensource project cannot be understated.
5. KVM is a bare metal hypervisor. The days of expensive “bare metal” software hypervisors like ESX are gone thanks to the chip makers commoditizing virtualization technology. It is even evident in the vmware stock price as industry analyst have come to realize that the landscape for the hypervisor is changing. KVM has access to the hardware and is a “bare metal” hypervisor just as ESX is contrary to what some may say. See blog from KVM maintainer.

With virtualization technologies becoming commoditized, the next growth area will be application stacks on top of the hypervisor. Some will be opensource and others will be proprietary which is fair enough as programmers need to eat too. (To be fair, vmware has always provided very good VM lifecycle management software on top of its hypervisors). The important thing for industry growth is to commoditize the right parts of the virtualization stack to which linux and kvm are playing its role. In my mind, I cannot see any other king of the hypervisor but KVM.