More On Private Clouds

Well, I seem to have entered an ongoing active discussion in the Cloud World with my recent post on "The Emergence of Private Clouds".  Even got a pickup from the press, which is always nice.

During the last few days, I've been discussing this with all sorts of people, and collecting the questions that are popping up.

I thought that it'd be helpful to recap the key questions -- and hopeful answers -- for everyone's benefit.

How Does This Compare ... ?

Understandably, there are some big cloud visions already out there from Amazon, Microsoft and others.  No surprise in that. 

How is this notion of a private cloud different?

I think it boils down to a few key attributes

First, there's the importance of preserving existing applications and information.  Most of the previous propositions meant using someone's stack to create or rewrite your applications.  I would offer (again) that waiting for most enterprise applications to be rewritten or recast is going to be a non-starter in many situations.

Second, there's the notion of control points.  IT is accountable for IT, and that responsibility doesn't go away simply because the source of IT infrastructure changes.  For the economics of the cloud to be successful, there are going to have to be clear control points for ensuring service delivery and security -- among other things. 

And finally, there needs to be notions of choice: choose what mix of internal and external resources you want to use, choose who your service providers might be -- and do so with a minimum of friction.  Take your applications and information wherever they need to go.

I'd offer that -- taken this way -- the current proposals really don't offer that sort of choice.

But What Am I Giving Up?

There are smart people who've listened to those paragraphs above, and have come back with "well, every approachhas tradeoffs, what are yours?"

I think that's a very fair question indeed!

First, we're presuming that the majority of the relavent IT landscape will need to be containerized using virtualization principles.  Without some notion of a virtual data center operating system underlying all the legacy apps and operating systems (and underlying infrastructure technology optimized for this environment), this stuff just won't work.

My view is that -- although there are some people a bit uncomfortable with this near-100% virtualization world view, we're much closer to this than many people think.  But, clearly, unless you're comfortable that this containerization will eventually happen, this view is not for you.

Second, this largely implies a world of x64 instruction sets -- basically, a cage match between Intel and AMD for the microprocessor of the future.  Why?  That's where are the significant virtualization work is happening in the industry -- and where most of the applications run today.

Third, this implies a progressive investment by IT customers in newer forms of tool to control and secure these fully virtualized environments.  I would argue that this is going to happen anyway -- private clouds or otherwise.

Doesn't It Make Sense To Just Rewrite Every Application Using (Fill In The Blank)?

This will be a healthy debate indeed.  As a technologist, I can certainly appreciate the beauty of rewriting our applications to use composed web services and RESTful protocols and modern frameworks.  Indeed, I do.

But those of you who actually work in larger IT environments will probably realize (as do I) that -- despite the intellectual attractiveness of the proposition -- corporate IT environments will unlikely muster the required investment to do so anytime soon. 

And that puts the anticipated business benefits out a very long way indeed -- as compared to other more pragmatic approaches.

What Does This Mean To IT Vendors?

A lot -- if you think about it.  If you follow the IT industry like I do, it's pretty clear that many vendor directions are lining up around these concepts.

Speaking from strictly EMC's portfolio, you can see the themes in storage, management and security being reflected in this line of thinking.

If you get very close to what VMware is doing, you'll see that they're way ahead of most of the industry in this sort of thinking.  Again, no surprise.

And recent announcements from Cisco and others point to a world where IT vendors are thinking about delivering the economics of cloud scale, but doing so in a way the provides IT the control they need, as well as a logical progression.  Hence the notion of a private cloud.

There's more than meets the eye -- and I'll do my best to illustrate more connective threads in the near future.

What Does This Mean To Enterprise IT Consumers?

Well, if you're already doing the virtualization thing, you're right on target.  You now have the option of looking at widespread virtualization not as an end-state, but as the first major step to building your own private cloud.

If you're not doing virtualization, or only thinking about it as a tactical tool in small portions of your landscape, maybe you'll want to reconsider your views in the future.  You're probably doing some of the right things, but there's an entirely different scenario you might want to consider.

More importantly, I think that -- once the industry likely understands and consolidates around these concepts -- there'll be a horse race by progressive IT organizations to move as quickly as possible in this direction -- the business advantages will be just too compelling.

When Will This All Happen?

Well, if you consider virtualization of IT resources and containerizing applications and their information as the first major step in the formation of private clouds -- well, it's happening today, isn't it?  But, clearly, there's much more to come here.

If we consider the second major step is "newer frameworks for provisioning, controlling, billing, securing etc."virtualized resources, you're seeing a flurry of activity from the relevant vendors in this category -- EMC being no exception.

And, if we agree that the third phase is "federated service providers that provide customers choice", well, it's early days indeed.  Near as I can tell, there are only a few service providers who've built environments from the ground up to received virtualized applications and informations -- and provide back to IT the control they need. 

More will likely follow, if I'm right.

How Can You Call Internal Resources A Cloud?

There are purists out there who consider cloud synonymous with the sole use of external IT resources.  Almost as if a cloud was defined by a physical location -- or, more precisely, a lack of physical location.

I consider that view far too restrictive, and utterly unpragmatic.

We all know of IT organizations that have sufficient internal scale to think in terms of internal (hence private) clouds that are more dynamic, flexible and cost-effective than what they're doing today. 

Building these newer elastic resource pools that use existing applications and information must certainly qualify as some form of a private cloud -- at least, to my way of thinking.

Harking back to an earlier point, IT organizations will more readily adopt cloud concepts when they're presented with flexible choices -- and one of those choices must be the ability to choose the desired mix between internal and external resources.

I'm hoping that we can evolve the cloud definition towards more of a behavioral definition, rather than one based on location :-)

Where Does That Leave Us?

In the early days of a fascinating discussion, if you ask me.

Clouds are cool -- but the vast majority of IT spend is with larger corporations running traditional applications. 

The real debate here isn't in terminology or semantics -- it's answering the question on everyone's mind: how will traditional IT organizations embrace the cloud?

And -- from where I sit -- the notion of a private cloud is the best model so far that I've seen.

But -- as always -- I'd love to hear your thoughts!

Comments

[Jay Fry]

It is indeed early days. And, yep, there's no way (either in a recession or not) that IT's going to rewrite all that they have for the cloud. So, an internal cloud's going to require you to be able to use what you have...plus a tool like what you describe to control what's going on. But, instead of requiring us to reach a world where everything is virtualized, that tool you talk about needs to be able to do the types of things you can do in a virtualized world, but act upon all of your IT resources, regardless of whether a resource is physical or virtual.

And, I agree: the interesting part of this is watching how IT orgs are picking up on this. Will continue to update you on what we hear from our (Cassatt) customers.