At 3:30 this morning, as I fed my ten-day-old son and dreamt of lost sleep, it occurred to me that managing a cloud can be compared to managing an infant.  Mind you, I’m not making a direct link between infantile behavior and those who work in a data center, but having experienced firsthand what it takes to roll out a private cloud, the process is much like the care of a new child. 

With newborns, the life of parents can be bewildering and even a bit cruel on the psyche.  There are a lot of requirements streaming in and, for the uninitiated, a big learning curve.  And if you fail to satisfy the demands of your progeny, a lot of screaming and general fussing ensues.  But, after some time goes by, parents learn enough about the needs and moods of their child to begin to anticipate those needs and moods.  And, by using this quintessentially human capability, we can begin to take steps to minimize the crying and sleepless nights.

But, in a cloud environment, it is software and not always people who must learn the personality of their charges – servers, applications, storage, and so on.  The raising of a cloud requires IT solutions to mimic the anticipatory function of man, and to translate this into an action or a resolution that makes sense for the environment. 

So, why is this important?  Imagine an environment with a distributed application that periodically dips in performance or even causes a crash due to heavy load.  Without analytics, your cloud is like a thermostat, responding to events that have already occurred like a dip in temperature prompts the heat to turn on.  You don’t actually avoid the performance hit.  Now consider what you could achieve if your system were proactive instead of reactive?  What if it could look at environmental data over time and begin to predict that the resource overload would happen at particular times of the day, or in particular components of the application, and then take steps to provide more capacity at those specific points to preempt the problem?

The ideal cloud is self-governing and efficiently automated, which is not possible without anticipation.  In fact, managing a cloud is a lot like managing many other things for which anticipation is a key ingredient for success.  Like an excellent butler, or iTunes Genius, or even Furby – that chattering relic of toys gone by – a cloud management solution must learn from its surroundings and become smarter over time so that its capability for anticipating the needs of the cloud environment is real-time.

Any cloud management solution, therefore, should include strong analytics with ability to correlate historical data with future events.  If you’re working toward true autonomy in the cloud, predictive, or anticipatory, capability of such a tool will soothe even the most fussy of clouds.

RedHat generated much heated discussion on their virtualization strategy when they initially announced their plans to replace their Xen-based hypervisor platform with RedHat KVM.  In fact, before the KVM product was even officially released, the anticipated switch was often a source of angst within both the RedHat Xen user community and the broader virtualization management world.  In cursory sweeps of random blog postings and short discussions with some of our customer base, I saw the technical merits of Xen touted over KVM, concerns about perforrmance, and open questions around platform management.

To be fair, there were also plenty of supporters.  But surprisingly, many of those in our customer base who initially questioned the merits of KVM are now taking a serious look at adopting RedHat KVM.  I’ve even heard from organizations that had initially rolled out RedHat Xen in their infrastructure but are now considering a move to KVM.

Why the seeming change of heart?  What is the rationale behind implementing this relatively new platform?

In at least two of the large customers in our base that represent major industries, the primary impetus for conducting a thorough analysis of RedHat KVM was their organizational initiative behind implementing a private cloud.  With cloud use cases at the top of these organizations’ requirements list, RedHat KVM provides an attractive, slim, and yet still easily-managed virtualization platform.

The most common reasons I have heard are:

• RedHat KVM is less costly than other virtualization solutions, whether for replacement or new rollout.
• Organizations with skilled Linux administrators already in-house do not have to make heavy investments in subject matter experts to make the leap to RedHat KVM.
• The ease of use oft-touted by users of Linux as well as the prevalence of the distribution, and its reputation for reliability make RedHat KVM an attractive choice, especially with its management layer capabilities.

Add to all of this the fact that KVM can host Linux and Windows guests, and you have a strong candidate for a private cloud platform.  Interestingly, one of those aforementioned large organizations has given the platform a production thumbs-up, and they are moving ahead with deployment.

So, is RedHat KVM the choice of private cloud platforms?  Frankly, it’s too early to tell, but I would argue that KVM is already proving itself in the virtualization space, despite the initial pooh-poohing of KVM by Xen proponents.  Organizations looking for an efficient and easy-to-manage cloud platform would do well to consider it.