Enterprise Cloud Wars

The recent thoughts about how "private clouds" might be the emergent model for enterprise IT has sparked some strong philosophical debate in the cloud-o-sphere.

On one side, the purists.  On the other side, the pragmatists.

Where you stand probably has much to do with where you sit, no?

Where We All Seem To Agree

Both sides appear to agree that we'll see cloud-like models in the future: elastic computing resources that minimize the need for traditional IT provisioning and management.

Both sides agree that -- over time -- the concept of geography -- where computing gets done -- needs to start to work in our favor rather than against us.

And then the paths diverge ...

The Private Cloud View

This view, promoted by EMC, Gartner, VMware and a few others, presumes that traditional enterprise IT can -- and should -- be part of the evolutionary model towards clouds.  Even Microsoft appears to be leaning in this direction as of late.

You'll hear terms like "hybrid models" (mixes of internal and external resources), "federated models" (again, dynamic mixing of resources that are owned by IT with those that may be rented), "trust boundaries" (used to signify the computing resources that IT directly controls) and the like.

Why do so many of us like this particular cloud model for enterprise IT?

First, it can preserve the enormous investment IT has made in applications and information.  Conversely, any cloud approach that presumes a rewrite of the entire application portfolio (or shifting to a SaaS provider) will take decades to get there.

As an example, anyone want to review the progress of SOA in our industry over the last decade?

Second, in a world of formal and informal rules and regulations regarding information, it's hard to imagine a world where any application and its information -- no matter how sensitive -- can run in a generic cloud in a generic location.  That's just being unrealistic.

Third, IT enterprises themselves can be efficient providers of IT resources using cloud models.  They have access to the same technology and processes as anyone else.  Many of these larger enterprises have considerable scale. 

And, unlike external providers, they've got relatively loyal customers at hand.

The Pure External View

An excellent paper was issued recently "Above The Clouds: A Berkeley View Of Cloud Computing".  The paper makes several excellent observations on what needs to change to make cloud computing widely successful.

However, it takes one rather extreme position -- it states the opinion that it can't really be called "cloud" unless it's completely external to the enterprise.  And therein lies the rub.

A few of us see this as an unfortunate (and divisive) view that will hamper the evolution of enterprise IT.  James Urqhuart, in particular, provides an excellent perspective on the Berkeley paper.

My problem?  The Berkeley academic view doesn't match other examples of industrial infrastructure we see in the world today.

Other Examples Of Industrial Infrastructure

When it comes to power generation, as enterprises, we're free to mix power from the external grid with power we may generate on our own.  Different businesses choose different combinations at different times.  Same goes for water.

When it comes to data networking and telecommunications, we see hybridized models everywhere.  Most businesses mix internal and external resources, and have the freedom to work with a wide variety of external providers.

The same discussion goes for supply chains, distribution, manufacturing -- as a matter of fact, it's very hard to come up with *any* example of industrial infrastructure that is purely external in nature.

Is enterprise IT all that different?

So, Where Does That Leave Us?

Given the definitions of "private cloud" that I and others are using, I've begun to see enterprise IT shops start to think in those terms.

Sure, the technology is different (e.g. fully virtualized environments), but what's really different is the dynamic, just-in-time management style they're using in these environments.

In terms of sheer adoption rates, I think these examples that I've personally seen far outweigh any large corporation using, say Amazon, for business-critical computing.

The problem with this particular purist model being proposed by the Berkeley academic team and others is that it leaves enterprise IT with unpleasant and unrealistic choices: forced migration of all applications, immediate adoption of new management stacks, forgoing of all internal IT infrastructure, etc. -- as well as the reality of a relatively anemic marketplace of external providers to work with.

Markets define terms, and I think in this case we'll need another year or so to see which perspective ends up being the dominant one.

What do you think?

Comments

[Tom Bittman]

Chuck, excellent post. You know what I think. Great contribution to the discussion!

In the end, we need to be realists. If we could all start over, we would design very different architectures. We can't. Evolution is messier. And when we're talking about IT, evolution is central.

The basic question is going to be - what is the most efficient and effective path for any given enterprise to leverage the cloud? For some companies (especially smaller), that will mean a "big switch". For others it will be a surgical switch, a hybrid switch, and an evolution. There are more variables occurring in this high-tech IT world that will continue to unbalance the equation. IT is not a fixed-price, fixed-function commodity - and services will also emerge, evolve, disappear.

[Chuck Hollis]

Hi Tom -- thanks for leading this discussion thread, and for dropping by.

I, like you, am far too pragmatic to imagine any "do over" when it comes to enterprise IT architectures, hence my philosophical disagreement with the cloud purists.

At the end of the day, what matters is what works, regardless of what label we decide to slap on it.

Thanks again!

-- Chuck

[Christofer Hoff]

Chuck:

A Great post yet again. I gave in and added my $0.02 with what I hope is a reasonably balanced view but with the thought that perhaps all of us who are commenting on the RAD Lab paper, weren't actually the intended audience in the first place...

If you've got a spare cycle or two, let me know what you think:

http://rationalsecurity.typepad.com/blog/2009/02/berkeley-rad-lab-cloud-computing-paper-above-the-clouds-or-in-the-sand.html

Thanks,

/Hoff

[Chuck Hollis]

Hoff -- I saw your material, and you made an excellent point, e.g. consider the intended audience. And, taken that way, it's pretty good!

The problem arises in the enterprise IT space, where everyone is looking for a little direction these days.

Thanks for taking the time to comment!

BTW -- check out Hoff's blog -- a man after my own heart!

-- Chuck

[irldexter]

When I mentioned interoperability, walled gardens and accessibility back in 2008 (in Melbourne in a large'ish boardroom of a larg'ish Telco) I never thought I'd build this: http://tr.im/gEMO

It hopefully will make you smile -> as I do actually support abstraction and virtualisation, just when and how appropriate, as per your blog post....

Keep up the good work!

[Chuck Hollis]

Too funny!! Thanks for sharing!!