in Musings, X-Geek

Virtualization and the death of KVM

I just got back from a client visit which lead to an interesting revelation. I went there to support the network management software that my company makes. Upon arriving at the client’s desk, I happened to notice a familiar KVM appliance sitting on his desk.

“Hey, I happen to know a lot about that KVM,” I said, nonchalantly.

“Yeah, I’m finding myself using it less and less,” he remarked. That’s when I noticed the KVM was disconnected. This gentleman happens to be the VMWare expert at this certain company.

His comment was telling, and I surprised myself in not thinking about it before. I can’t recall a single customer I’ve visited in the past year who didn’t have some servers running VMWare. I run into it literally everywhere. This fanatical adoption of VMWare made me prod my broker to buy VMWare stock. Like I predicted, VMWare had an outstanding IPO.

Virtualizing servers cuts down on costs. Hardware costs, especially. Fewer physical servers are needed running VM sessions. Dell and other vendors will still sell hardware, particularly hardware optimized for virtualization. Software costs, on the other hand, are roughly the same. You still need your Windows licenses for Windows. In fact, Windows’s instability and vulnerability to viruses helps drive virtualization efforts, since the admin can now simply reimage an errant server. That means Bill Gates can breathe easy about virtualization.

The biggest loser against virtualization is the KVM vendor. Every new virtualized server means one less KVM port. If new servers are being totally virtualized – and at many datacenters I’ve visited this is exactly what is taking place – the KVM business is dead, pure and simple.

Virtualization is one of the fastest technology adoptions I’ve ever seen. If your livelihood depends on the KVM business, you’d better find a different business, pronto.

  1. A few years ago I attended a VMWare seminar. They had a speaker from Stock, the large wholesale construction supplies company. They grow by acquiring smaller companies with an established business. The result was that they were drowning in application servers. They showed a before picture consisting of a basketball court sized server room without enough room to swing a cat because of all the various servers and racks all over the place.

    The after picture consisted of one large hp rack system. Stock installed the native vm system that runs atop the intel hardware directly. The systems have the ability to statefully reconfigure any of the systems from a single console. They can grow or contract a systems resources at will. It was very impressive.

  2. We had a hyperviser that we used in 1970 to run multiple OS/360 systems on the same hardware. It allowed us to have multiple test paths on our systems development. Not even IBM of those years could afford a standalone 360 for every development group. The program, CP/67 had to run in a system with time slicing and dynamic address translation.

    Incidentally, it also gave a method to test new computer instructions before the hardware existed. You had your assembler generate the code for the new instruction, even though the hardware didn’t support it. The code would “program check”; but, the hyperviser would “catch” the error, simulate the instruction’s function and give control back to the program at the next sequential instruction.

  3. To steal Mark Twain’s famous phrase, the death of KVM has been greatly exaggerated. If the suggestion that I should find a different business pronto because server virtualisation will kill off the need for KVM switches then the same must be said for manufacturers of servers, network switches, racks, Redundant PSUs and anybody else involved with the supply of product for datacenters.

    Server virtualisation allows for more to be done in less space, I can’t see anything new in that. Since the birth of digital computers in WWII (thank you Tommy Flowers) computing has constantly done more in less space.

    In 1943, Thomas Watson the then Chairman of IBM said “I think there is a world market for maybe five computers”. History has shown that as more powerful computers become available more is done with them.

    Server virtualisation allows more to be done with a box. That box still needs human interaction and that interaction will be done via a screen keyboard and mouse. Unless or until a new interface is invented then KVM switch technology will be required to manage servers, what ever is running on them.

    Having said all that, KVM technology is changing and adapting all the time. This year has seen the birth of the virtual KVM switch. Traditionally the KVM ports of a server were plugged into a KVM switch where a number of users, typically 4 – 8, can then access any of the attached servers. What happens if 9, 10, 15 or 50 users want to access any of their servers? Until recently the answer was to buy ever larger, more complex switches. The logical end point is an infinite number of users to an infinite number of servers. That will be a big switch! Alternatively, existing technology can be adapted to provide exactly this solution.

    KVM-over-IP has been around for some time but for the most part it has allowed users to access and control a KVM switch and the servers attached to it over the LAN or WAN. If you give each server its own KVM-over-IP device and provide a user friendly interface to select a KVM-over-IP device to be accessed and therefore which computer can be controlled, then the need for a KVM switch becomes redundant. The network becomes the KVM switch, or you have a virtual KVM switch.

    It is innovations like these which mean that KVM vendors such as Adder Technology (, who have just launched the AdderLink ipeps (IP-Engine-Per-Server) virtual KVM switching solution, have a very strong and healthy future regardless of Server Virtualisation.

  4. Hi Richard,

    Thanks for adding your comments. You argument doesn’t convince me but I do welcome your viewpoint.

    Since writing this post I have spoken with several more customers. Each one without fail has mentioned their aggressive embrace of virtualization.

    I stand by my statement that the KVM market is a dying one. Good luck, sir.


Comments are closed.


  • Mark Turner - A Life, Unfinished » Blog Archive » Brief MT.Net Outage September 11, 2007

    […] far I’m digging the service. Like I said previously, the days of the absolute need to run on bare hardware are now over. Funny how the early days of […]