Baby (Private) Clouds

I'm having a lot of fun using the concept of a "private cloud" to explain how IT infrastructure is likely to evolve in the near future.

So far, it's causing some interesting discussion -- not only with the customers I meet, but in the broader blogosphere.

And, once in a while, I get a wonderful surprise.

The Basic Ideas

You're welcome to read the explanatory post I put up a while back, or some of the Q+A that's come up already, but if you're too lazy to click and read, here's the summary:

• The current state of IT infrastructure can be charitably described as inefficient -- far too little of our current investment ends up in new business value, far too much is spent keeping the lights on.
   
• As a result, people are getting frustrated -- business people, and the IT people that support them.  Almost every new business initiative ends up being an IT project of one sort or another, right?
   
• And, right now, we're getting a lot of new thinking in the industry: clouds, virtualization, etc. -- but we need a consistent model that describes how enterprise IT will likely evolve, and why.
   
• A "private cloud" model melds the advantages of cloud (flexibility, economics, geographic independence, etc.) with the control (service delivery, security) and migration path (existing apps) that IT needs to run the business. 

By containerizing applications and information (think virtualization), it won't really matter what server they run on, what user device is being used -- or even who owns the infrastructure, if you'd like.

Described this way, private clouds are distinctly different than other enterprise cloud models being proposed in several important regards:

• They preserve the considerable investment in existing applications and information.  Rewrite if you'd like, but it's not needed.
   
• They provide the control that enterprise IT needs to run the business.  Without these control points, IT isn't doing their job.
     
• They enable a flexible and dynamic mix of internal and external infrastructure, depending on your needs and interests.
   
• They are achieved via a natural, logical and risk-managed migration path that doesn't require a big leap of faith.

Now, understandably, there are a few "cloud extremists" out there who'd argue that none of this is truly a cloud in its most pure form.  Fortunately, most people in our industry are far more pragmatic when it comes to enterprise IT.

So, You Were Going To Tell A Story?

Yes, I was -- sorry ...

About two weeks ago, I took 15 minutes to run through the basic concepts of a private cloud in front of one of our customers who were in for a briefing.  AS always, I'm looking for thoughts and feedback.

Imagine my delight when one of the senior IT leaders said "well, I guess we're already doing that today".

Turns out that during the summer of 2008, they'd stood up a large-ish VMware farm around some non-traditional thinking:

• The farm was positioned as a "free" service for anyone who wanted to run something but didn't feel they needed industrial-strength IT infrastructure just yet.
    
• They'd set up a simple web tool where anyone could request a limited number of standard build stacks, press a button and immediately get a login prompt -- no questions, no waiting, no approval.
   
• They'd done some lightweight automation on provisioning, service delivery monitoring and management as well as a few other aspects.  They were interested in the newer management tools for virtualized environments, and this gave them a great way to get some practical experience.
   
• If an application on the farm became more important, they'd take a look at more formal infrastructure requirements.  Conversely, if somone lost interest in the application, the virtual machine and its information was archived, freeing up the resource for someone else.

No surprise, this sort of platform turned out to be extremely popular with their user community.  It didn't take more than a few months before they'd pretty much filled up the storage (not the servers!) they'd allocated for the project.  I guess I'm OK with that :-)

More than a few apps had been graduated upwards to more formal infrastructure -- still virtual, but built and managed differently.  The vast majority of apps had sprung into existence, run for a while, and then were quietly put to sleep on cheap media where they could be retrieved quickly if needed.  Hence the storage requirement.

This person offered that they had built their own private cloud -- maybe not a full-fledged one -- but I'd agree that all the concepts were there in some form. 

Maybe call it a baby private cloud?

Other Thoughts

Of course, by now I was smitten.  I needed to know -- why'd you do this?  I mean, this sort of investment (and IT management style) must have been driven by something or another, right?

The answer was pretty logical:

• IT processes weren't keeping up with an ever-increasing river of small requests for one app or another.  None of them were individually big deals, but -- collectively -- there was a lot of work involved. 
   
• As a result, IT was starting to see much more "shadow IT" where someone stands up a server and storage in an office closet or cubicle.  So the company was spending money on this stuff, it just wasn't coming through IT.
    
• They knew they'd be doing far more with VMware in the future, and needed a test bed to try out new tools and processes -- one that wouldn't potentially crater an official production application.
   
• And, frankly, they thought it would be a really cool idea to do something like this, just to learn from the experience.  That one I really understood :-)

What They Saw, They Liked

Put in the context of the private cloud discussion, these people liked what they saw. 

They agreed with my assertion that -- yes -- most of their production environment (server and desktop) could be virtualized today, but the tools (and associated workflow processes) had some more maturing to do.

They liked the idea of containerizing their applications and information so that it was completely agnostic to whatever server or whatever user device was needed.

And they really liked the potential emergence of federated service providers where they could flex their IT infrastructure depending on economics and a variety of other factors.  Or not, as the case may be.

And then they wanted to see roadmaps, which -- of course -- I can't really share here, can I?  We ran out of time, but agreed we probably needed to reconvene and do some detailed roadmap reviews.

A Different Way Of Thinking About IT Infrastructure?

Lots of people are busily virtualzing their existing applications -- all good. 

What I liked about this story is that I saw a fundamental shift in how they were thinking about the problem: dynamic vs. static, automated vs. manual, flexible vs. partitioned -- kind of a "just-in-time" IT infrastructure approach.

Put differently, you're not really changing what you're doing -- providing IT resources -- you're just radically changing how you're thinking about the problem.

So, I'd like to hear from anyone out there if you've seen something similar?

And, if you have, can we consider these sorts of environments "nascent private clouds"?

Comments

[Shibin Zhang]

I started a kind of battle with Ms. Warrior:

http://blogs.cisco.com/news/comments/introducing_unified_computing_to_the_data_center/

because I still insist that the future is green IT and virtualization is just a sub-set of next generation.

Shibin

[Sudhir Brahma]

Hello Chuck,
As an avid user of "virtualization", I realize there is a dark side to it and am not sure why no one is talking about it: I can actually build a complete Virtual machine , install all the licensed software I want (just pay for it once and register)...and then this image file on the hard disk or whatever storage media is used, is ready for replication. Each replica has all the software I paid for once and it can be used by just about anyone I choose to give!!
I think we are getting close to the paradyme created by Xerox, when they created the photo-copying machines and had the book-publishers running for cover.
PS: I am typing this message out one one such cloned machine-still with me and not give out to anyone :-)!!
regards
Sudhir.brahma@gmail.com

[Chuck Hollis]

Interesting point -- maybe that's why so many software vendors are reluctant to embrace a virtualized world!

That being said, just about every enterprise I talk to wants to be 100% compliant on licensing -- regardless of technology.

Thanks!

[Sudhir Brahma]

They may still be 100% compliant with the licensing policy as it applies today- all of it depends on what the licensing policy itself is. If it is tied to a "machine" or a "drive", they may still be on the right side of the legal, though treading gingerly enough to be let off with a warning!
Besides, there is no law with says you cannot "share" your machine...it just happens that this process of "sharing" just gets extended since the machine itself can now be transmitted over the wires!!
So theoretically, I can "ship" the whole machine with MS office for someone to work on a word document!!...that so as long as I don't use it at the same time! But there is nothing which "enforces" this "use one at a time only".
I think this a Paradyme change and the licensing processes and the regulations have to adapt suitably (while still not being a nuisance for legitimate users).

regards
sudhir.brahma@gmail.com

[Lee Paul]

Chuck, your commentary on rethinking traditional business processes reminded me of a great book from grad school - "Blue Ocean Strategy" by Kim & Mauborgne. It promotes creating new market space or "blue oceans" rather than competing in an existing industry "red oceans".

In this context, the company your wrote about gained a competitive advantage by rethinking existing solutions to their business problems. This company is discovering how to move non-mission critical requests away from the system admins that support them. This saves time and resources by letting the admins focus on more important initiatives. Once a project gains momentum, the company can re-prioritize how they allocate storage space based on business need. Meanwhile, this company's competitors may still be operating in the traditional BAU process.

To Sudhir's licensing comments:

When a customer licenses applications on a registered client/device model, it makes no difference which persons access the device. Only that the device itself has the right to access the application.

Consider this example:
A corporation purchases a Bloomberg license on a dedicated desktop. Since the device itself was licensed, anyone in the company is allowed to access the application. Since it's just one desktop, only one person at any given time is physically able to use the machine.

Does Bloomberg care if the application is virtualized, and someone across the world accesses it? Not if only one person can access it's application at any given time.

Virtualization didn't provide the corporation with additional usage. (Other than not needing to buy a separate desktop license somewhere else in the world).

If the application was licensed to a specific user, virtualizing would still not be a factor, considering only that person was granted the right to use the application. Thus, someone across the world cannot use it in order to be compliant. To Chuck's point - corporations have become extremely wary about compliance, and almost always act in good faith.