Tuesday, June 12, 2012

A Step Between

Anyone who knows me or has read this blog understands my views on private cloud.  As I've often stated, the vast majority of private cloud implementations and solutions (in nearly all I know of) are really mass virtualization.  In essence they provide nothing in terms of services, tools, or process to support the application layer.  As a result enterprise cloud adoption in the Fortune 500 is slow because enterprise native cloud applications just are not being built today by the builders and owners of these private clouds.  Instead vast numbers of applications are being virtualized which is, of course, a good thing.  Lower costs, higher efficiency, and improved return on assets are lofty goals, and delivering on them certainly looks good to the CEO and CFO.

What I realize today is that perhaps we need a step between mass virtualization and the cloud, a new on-ramp encouraging companies to continue down the cloud road.  Private clouds are viewed by some as a step between virtualization and cloud; a path justifying a measured approach while focusing on how to leverage existing assets and move up the learning curve.  Not unreasonable and I'm on the record as to my concern about unintended consequences.  And one real, indisputable consequence is that private clouds have made virtualization a requirement and made cloud more palatable.  The proverbial ball has been moved forward.  However private cloud is just a limited version of cloud.  I feel we need something fundamental, a capability which fills the gap between operating system dependent machine virtualization and operating system independent cloud resources.  So the question is what should be the next step so we keep the ball rolling and continue gaining momentum?

For the foreseeable future applications will run on operating systems, although I'm also on the record on this topic as well.  Server virtualization enables us to load up applications inside a protected container, the operating system, separated from other applications by the hypervisor.  The hypervisor is an artificial construct created because applications simply do not share resources well, and the most common hypervisor, VMWare, is not cheap.  But how much value does a hypervisor provide? If the operating system were improved to provide a sandbox for each application to which resources could be applied; and tools were added to provide application management capabilities including images, backups, and portability; we would no longer need the hypervisor.  The grid computing pioneers, and specifically United Devices, had it right from the start.  They provided a sandbox in which the grid application would operate and to which resources could be assigned or shared.  Who says grid isn't the father of cloud?!!

Of course none of that matters except to the innovator.  What does matter is that software licensing, which fell behind the cloud curve as publishers struggle to figure out how to protect revenue streams and intellectual property, ends up nicely realigning in a container world.  Licensing of operating systems would be easier to manage, but more importantly more efficient because fewer instances should mean fewer administrators.  And an entire layer of software costs would be eliminated, a layer that can cost nearly as much as the server it runs on.  Based on my experience the cost of a cloud environment would drop by as much as 30% to 45%.  That's enough to interest anyone.

In this transformation from hypervisor to container, the operating system battleground could reverse direction, from low end operating systems such as Linux and Windows back toward data-center-centric UNIX stalwarts such as Solaris, AIX and HP-UX.  Oracle would be in the catbird seat because of Solaris Containers, that is assuming they know what they have.  However many companies have migrated off Solaris to Linux, as is happening to AIX And HP-UX.  So what is the core Linux team working on? What are the various distribution providers preparing to bring to market?  Of course a move by the Linux world would more likely than not drive Microsoft forward, but only after the concept is proven.  Naturally VMWare would have to reinvent itself, Citrix would have the opportunity to double down on CloudStack.  And the system management vendors would have to update their tools too.

The best news is that current cloud providers wouldn't need to change anything immediately.  Over time they would be able to seamlessly move to the new model and support a bifurcated environment in perpetuity if necessary.  Integration points would change but the right recipe for success is for the innovators to make the change invisible to the application and systems management tools.

I'll have to start digging to see if anyone is working in this direction.  It may be the new bandwagon I need to jump on as my current bandwagon, Enterprise Private PaaS, is a five year marathon.

1 comment:

  1. It almost feels like the founders of Docker read my post :)