Dell CTO Kettler on virtualization

Dell is apparently putting some of its muscle behind blogging.  This morning, I woke up to an email that Dell now has a manager -- Lionel Menchaca -- for its public facing Direct2Dell blog.  Apparently, the strategy is for any commentary coming from Dell execs and techs to be poured into one blog.  It's good that Dell is blogging but if it was me running the show, I'd establish individual blogs for each of the key Dell folk instead of one for all of them. If Sun CEO Jonathan Schwartz can do it (and now Sun general counsel Mike Dillon is in the action), it stands to reason that other vendor execs can make the time so we can get to know them better (instead of having their personalities diluted by one big bucket).  That said (and in all fairness), maybe it's just harder for the lieutenants to post frequently enough for us to get to know them. Since December of last year (2005), Sun CTO Greg Papadopoulos who at one point was posting at a decent clip has only published two blog entries; one in March and the other in June.  Greg may have been swallowed whole by his new duties since Jonathan took over.

Speaking of CTOs (and getting back to Dell's blog), the most recent entry on Direct2Dell comes from Dell CTO Kevin Kettler who went both the text and video routes, the latter of which is a whiteboard session. By the way, can anybody tell me who the most prolific publisher of whiteboard videos (scroll to the bottom) has been for almost two years?. Thanks Dell (on behalf of Marianne Wilman who drove the idea internally here at CNET). Imitation is definitely the sincerest form of flattery. Anyway, Kevin's latest entry explores virtualization (one of my favorite topics). Under the headline Virtualization, what's the big deal?, Kevin wrote:

Perhaps the biggest change will be in the lower cost of implementing a virtualized system. Virtualization software is more and more common—some is even free open-source. Moving forward, this means two things: lower cost of implementation, and more software options for customers... both of which are good for virtualization.

Virtualization is a complicated topic. I'm not sure what Kevin meant by more software options. But when you think about software "stacks," virtualization is turning out to be a stack of software unto itself (sort of a stack wihin the bigger stack). Today, there's the hypervisor, host OSes, guest OSes, and a few management utilities (basically, interfaces to control virtual machines). Over time, the management layer will get broken down as innovators come forward with highly specialized software that does one or two things really well. The stack will grow to include paravirtualized operating systems (operating systems that are aware of the fact that they're running in a VM, and that autonomically self-optimize for that enviroment) as well as specialized OSes (practically applicances) designed to do nothing else but proxy hardware drivers as a host to the guest OSes.

So, overall, I agree that there will be more software options. But within certain layers of the stack, for example, at the hypervisor layer, I actually expect the industry to consolidate around certain standards which may eventually mean fewer software options (at certain layers). 

On the cost front, I don't buy the idea that open source translates into lower cost of implementation. Well, maybe it does, depending on what you count as implementation.  We hear a lot about this question and usually, the reconciliation includes some mention of how it all comes out in the wash when you consider the business models behind open source (subscribing to support) vs. the licensing of commercial software.  Virtualization software is not simple stuff and as the stacks get more complicated, so too will the skillsets of the people in charge of running them. The smartest thing I've heard on the issue of going the open source route came from Mary Hodder who recently started up Dabble. In my podcast interview of her, when the question of her startup technology came up, Mary told me:

What happens with open source is you actually spend the same amount of money, but you don't have lock-in and you pay for really good people to run it.  And so you still end up paying. But you just pay in a different place. And I think it's a much more sustainable model for that kind of server/software development.

Her thoughts inspired me to post a completely separate blog about the opportunity costs (great people to run your software vs. software that supposedly runs itself). Again, this comes back to what you count as "implementation." Kevin goes on to say that:

Virtualization will not be a passing fad. What will be interesting in the coming years is to watch how software licensing will need to change and how virtualization will lead to new ways to package and distribute software.

Amen Kev. I've written many times that resisting the benefits of virtualization will be like resisting gravity. Even on the client side.  And there will have to be a revolution in licensing because no one is going to put up with the current models where you may end up having to pay for multiple instances of your OSes and applications (one instance per virtual machine) even though you're the only one using them. Finally, one of Kevin's other interesting points:

With virtualization, a software developer can do some pretty creative applications in the contained virtual environment. This is all pretty exciting stuff and it’s got us brainstorming new ways that virtualization can solve problems and advance the industry in both the client and the server. I'd like to hear where you think this is going.

Well, how about this? Dell can think about taking us there, particularly on the client side. How about delivering Dell desktop systems pre-virtualized and then come up with a management utility so that when the time comes to move to a new system, it's as simple as copying a virtual machine from the old system to the new system (and taking care of all that complicated licensing management stuff caused by Microsoft's Windows Genuine Advantage program). 

Upgrading to new systems is one of the most painful things you can possibly do. If you're a vendor of desktop/notebook systems, it also represents that point where you can keep or lose a customer.  Today, most system vendors have pretty much nothing from a technology point of view that "encourages" loyalty. Upgrading from an old Dell to a new Dell is no easier than upgrading to a system from a competing vendor. The system vendor that figures out how to make it less painful to upgrade to their own systems than to their competitors' is the one that will get more loyalty out of their customers. Virtualization is the easiest way to tackle that problem.