virtual machine

I’m a serial entrepreneurial leader.  It’s an art/science, left/right brain thing. I have to say that one of the most challenging parts of creating a compelling strategy, leading a company or building products is getting people to see the possibilities, transitions and tipping points. Imagineering the future calls me to look back at what made companies great -- specifically, how they capitalized on paradigm shifts while the rest missed it. Reading the recent bestseller, Outliers, it struck me that, not only do you have to be smart, but you have to be in the right place with the experience to see and grab the brass ring.

Moore’s Law is one of those history lessons that have traditionally been a touchpoint that points the way to the future. Simply put, Moore's law describes a long-term trend in the history of computing hardware, in which the number of transistors that can be placed inexpensively on an integrated circuit has doubled approximately every two years. 

Translation:  compute power has reliably doubled at a decreased cost every two years.

In a recent announcement, Intel gave a glimpse of what the future will look like. The “Cloud” chip will have 48 cores, is available to Intel’s ISV partner today and will be shipping in volume in less then 18 months. The quote from the Intel dude stated that it will increase the power of what is available today by 10-20 times.  Oh my…. Buckle your seatbelt …. Moore’s law just took a giant step up the paradigm.

In one of my discussions with some folks from VMware I have heard pretty much a uniform response that the tipping point for VM adoption in the data center was the introduction of dual processor chips in 2007. Two core CPUs + VM isolation means I can consolidate physical boxes onto 1 machine. How simple is that math?

Gartner says that the average number of VMs in the data center per CPU socket is 10+ or 5 per CPU.  VM Density is how many environments can run on CPU socket. VM Density increase is due to better through put of the CPU and more efficient VMs and is “the new measure of IT efficiency.”

Fast forward to the data center of 2012. For a moment lets ignore other bottlenecks in the stack that might stop us from drawing a straight line from today to then. VM density will be 240 (5 * 48) per CPU socket. Think of all the empty space in those data centers. Certainly enough to store a few hardcopy versions of the US’ accounts payable to China.

What are some other data points that we can look at today that can help us see the future more clearly? Today’s VM landscape breaks down close to 80% Windows, 15% Linux and 5% other. Why is there such a high concentration of Windows in the data center? Not that I enjoy poking a sharp stick in the eye of Microsoft, but I have to say that it is because no one that wants to keep there job will run more than one application on top of  a Windows 2003 or a Windows 2008 server.

Running multiple applications on Linux? No problem. It is this 1 to 1 OS to app ratio that is one of the things that has made VM adoption so compelling. Fix the fragile Microsoft OS by using a hypervisor to create the isolation between applications that Linux has out of the box is sweeping the industry. So Windows app density is high in Linux, and 1 for on Windows server.

Let’s take a use case where all the VMs are Windows based in the data center. That means there will be 240 copies of the same operating system (assuming the market has actually adopted Windows 2008 server by then). There has to be a more efficient way to deal with the fragility of Windows than running so many copies of the same thing.

Enter application isolation technology or Virtual Application Appliance (VAA) for server side apps. A VAA is a container (a cloud container?) that isolates an application from the OS and other applications. This isolation makes Windows de facto more reliable and eliminates the challenges that force the 1 to 1 App to OS deployment design pattern. VAA makes Windows deployments as robust as Linux, increasing the app density on Windows. Who wouldn’t want that?

Now I am not suggesting that it would be a good idea to run all 240 apps on one Windows 2008 Server OS, increasing the app ratio from 1 to 240. But running 6 applications on 1 OS?  That is very doable and it will reduce the number of concurrent OS from 240 to 40.

Think of the memory savings (recommended memory configuration is 2 GB per OS or 480 GB of memory in this use case). Think of how much you could shrink the checks you have to write from Microsoft by increasing the app density ratio from 1 to 6. An 84% reduction in #OS running. Intuitively, you know it's cheaper to run multiple apps on one OS, but how much cheaper? Well, Amazon is expert at pricing cloud services, and on a simple example of 3 VMs running Windows with 1 app in each VM, the monthly cost would be about $260 on a standard small instance. Therefore running on Amazon with one VM running Windows and 3 apps, the price is $86. Get the picture?  Less OS licenses, less VM licenses, less CPU cycles, less disk, less management

So here are my predictions for 2012:

• Intel cloud chip will be shipping a 48 core piece of silicon for less then $500
• VM density will exceed 200 on this chip in the data center
• Windows OS will be over 75% of the VMs that run in the data center
• 200 copies of same the Windows operating system running on 1 chip is silly
• The industry over time removes silly inefficient execution stacks
• Application isolation will have crossed the chasm addressing, even eliminating, this inefficiency

Assuming the Mayans are wrong and there will be life after December 21^st, 2012, you can see my take on the market.  I would love to hear how others see the impact of the cloud chip on virtualization and cloud computing. Greg Ness how will it affect the infrastructure 2.0?  James Urquhart impact to the Wisdom of clouds is just around the corner?  What is the next tipping point? Drop me a line at GregO {@} Appzero {dot} com or tweet me at http://twitter.com/gregoryjoconnor.