Private Clouds And The Fixed Vs. Variable Discussion

Yes, I'm still talking with customers about private cloud models, day in and day out.  The funny part is that now they're asking me, rather than me bringing it up.  That's progress, I guess.

Just to refresh your memory, in my simple world clouds have three major attributes:

* they're built differently -- dynamic pools of virtualized resources

* they're operated differently -- zero-touch or low-touch aggregated models

* they're consumed differently -- they're convenient to consume

I seemed to have spent the first part of 2009 discussing the technology.  Fortunately, many people now generally accept that fully virtualized environments can be better than their physical counterparts in many situations.

And, as I think about it, I spent a great deal of time during the second part of the year on operational models: the need for self-service, the importance of flexibility, how traditional roles and responsibilities change, and attacking the people / process / politics part of the problem.

Haven't solved that one entirely yet in people's minds, but we're starting to make progress :-)

Which brings us to the third chapter: different consumption models.  And therein lies the most critical debate: how is all this great stuff paid for?

The Classic Shared Infrastructure Debate

I bet you could run an entire Econ 307 course on all the different models on how shared infrastructure is paid for in our world. 

Highways, for example, are ostensibly paid for by taxes.  In the US, we are told that fuel taxes pay for the roads, although I'm suspicious.  I also seem to pay a lot for tolls as well.  And, if I drive a bit too fast, I pay another form of tax for "infrastructure support services", e.g. the cops.

Different models for air traffic control, telephony, power, etc. etc.  Paying for shared infrastructure is not a hard concept in our world -- there's plenty of it, and it seems to be paid for in some degree.

Even within enterprise IT, there's plenty of shared technology infrastructure that doesn't seem to cause heartburn.  We all have access to the corporate data network, as well as shared services such as email and other productivity applications.  No real drama here.

And, if we take a *really* macro view, all the IT investments come out of the same uber-enterprise bucket at the end of the day.  We're really talking more about the intricacies of internal cost allocation in most cases.

The Fixed vs. Variable IT Conundrum

There is a sizable portion of the IT landscape that hums along in a predictable manner -- workloads are understood and predictable, the application and management stack is in place and works acceptably, etc.  I visualize these environments like power generators sitting in the basement.

Going to these environments and saying "hey, you need to fully virtualize your environment!" is not an easy sell.  The legacy stuff works, and it's paid off -- for one thing!  When it comes up for the inevitable refresh, we'll make our case then.

Sure, these sorts of IT application / infrastructure stacks might get some benefit from pooling of resources (assuming workload demands are uncorrelated!) and there might be some nice benefits in migrations to more efficient kit, or workload balancing, or a better way of doing disaster recovery, etc. -- but, in the big scheme of things, these are usually nice-but-not-game-changing, get-you-promoted benefits.

On the other hand, it's fair to say that a lot of interesting IT is essentially variable in nature.  Some workloads are inherently variable -- dev and test comes to mind, as does decision support and business analytics, not to mention the new wave of interest in virtual desktops.  Surge-y, burst-y stuff that usually isn't exactly bet-your-job mission-critical.

Sure, the case for dynamically pooling resources makes all sorts sense there on a lot of levels.  Physically overprovisioning these workloads to "theoretical worst case" is just plain wrong.  And I think most people understand this.

But there's a very important use case for "variable infrastructure" that I'm finding that many IT organizations are missing.

I'm talking the new project, the new application, the new use case, the shiny new thing that the business wants to do.  The workloads aren't fully understood or are highly variable.  The application and management stack might be changing.   No one is quite sure what "steady state" might look like, or (more typically) if it will ever be achieved.

In my travels, I sometimes meet IT groups who are almost completely focused on "fixed IT".  They are laser-focused on availability stats, efficiency of resources utilitized, etc. etc.  Change is not their friend; they do everything in their power to minimize not only changes in requirements, but the impact of changes in their requirements.

And, if you're in this world, my hat's off to you -- you've got a job to do, and you most likely do it well.

However, I'm going to argue that more business value can be potentially created on the variable side of the IT equation.  And, if you do IT strategy, and you're not trying to come up with better ways to handle the non-optimized, not-predictable, not-poured-in-concrete side of the equation, you might be missing something big.

Here's my case ...

Creating Business Value With Variable IT Resources

If you're reading this, and you've ever been on the business side, this might sound familiar. 

You are a business person with a problem or an opportunity.  You want to try something new.  To do this new thing, you'll need IT infrastructure resources of some sort.  You're not quite sure about how many, how fast, how secure, how big, etc. etc.  It's early days.

What you'd like to do is to be able to try something quickly, see how it works, and then adapt quickly as new needs emerge.  Maybe it's a new application, or a new mashup of information, or something else.

You, the business person, go over and talk to your IT guys.  They (often) want to know (a) how big/important/complex this will eventually be, (b) how much money you have, and (c) how long you're willing to wait for something to go use.

Your answers?

Typically (a) I don't know, (b) not enough and (c) I need something right now. 

Now, I'm not casting aspersions here, but I would be surprised if a traditional IT organization wasn't 100% responsive to this sort of business request.  After all, traditional IT is focusing on the Really Important Applications that Run The Business.

Now, if you're in a competitive environment (who isn't?) and you have a fair amount of smart knowledge workers, answer me this -- where does business innovation typically come from?  And how can a smart IT infrastructure strategy support this flavor of business innovation?

Indeed, I've seen many organizations where this is the domain of "shadow IT" -- bootleg servers, side deals with outside service providers, free web resources, data closets, etc. -- stuff that's deliberately kept off of the formal IT organization's radar screen.  It's out there, and we all know it.

Now, imagine if you had a bunch of on-demand virtual machines that you could make available through a simple self-service portal, and you -- gasp -- turned them over to just about anyone who had a valid cost center.

I'd bet that self-service vending machine for virtual servers would be mighty popular.  Business users could get what they needed quickly, and adjust their requirements without too much fuss.  The business would spend money on this stuff far more effectively than people doing it under the covers.  Prototypes and experiments that actually gained traction could be graduated to more rigorous environments.

And, best of all, it'd be easy to keep a watchful eye on all of this.

But how to pay for it?

The Biggest Obstacle

Most IT people see a big Vblock-type of thing, get it and want it.  If they're thinking about an existing VMware farm, one type of thinking kicks in, e.g. improvements over my existing approach.  But usually, they're thinking about unmet needs in the business -- that's a good thing.  Perhaps some sort of private cloud with a self-service portal.

And about 30 seconds later they go "oh my gosh, that's a big rack, how are we going to pay for / justify / rationalize this?".  Most of the relavent budget is usually tied up in the legacy environment.  And no one's coming forward with a big pile of money to pay for the shiny thing.

So -- even though the need is recognized, and the technology model is viable -- how do you get started on justifying this sort of self-service environment?

There are some really clever people in our industry, and here's some examples ...

One person I met went to his purchasing department, and found out how much money was being spent on beefy servers, storage and desktop stuff outside of IT.  He added it all up (it was a very big number) and smaller Vblock-type thing looked like short money in comparison.  Made a case to the CFO that there was a better way to spend the company's money.  Done.

Another person went and met with his crankiest user group -- the ones that were always dissing IT for being slow, unresponsive, etc.  In this case, it was a department of analysts doing heavy-duty financial modeling.  He pitched what he wanted to do -- a shared compute farm, controlled by them -- and asked for some initial funding with some obvious skin-in-the-game from IT.  He got it.

A third found an internal IT use -- test and development of non-mission-critical applications -- and simply put together a compelling case based on internal IT spend!  Didn't have to venture too far outside of the data center to see that one.  Smart.

There are many more examples I could share -- but the pattern is clear: think about the problem differently, and the money is usually there (albeit being spent elsewhere!) to go do what you think needs to be done.

The Duality Of Fixed Vs. Variable Consumption

If you're modestly engaged in physics, you know that matter and energy are two sides of the same coin.  And, if you think about fixed vs. variable consumption models, the same thing is true.

Fixed costing is nothing more than pre-paid variable consumption.  If you don't believe me, look at your cell-phone bill, especially if you've got one of those "bucket of minutes" plans, or -- better -- unlimited access.  If you lease a car, the same thing is true -- especially if you exceed your mileage cap!

In one sense, variable pricing can often be thought of as nothing more as an on-ramp that leads to a more predictable costing model in most situations.

Put differently, unless you can offer your business users a convenient capability that's priced to their budget, they'll end up going somewhere else.  That's what capitalism is all about, and most people on the business side are, well, capitalists!

And, ultimately, forcing them to do stuff off-the-radar -- that's not good for IT, and not good for the business.

Comments

[Sidhartha]

Your views are great and future seems to be bright for Cloud computing. One point I like to add this technology will remove the energy consuming hardware.