Untangling Enterprise IT For The Cloud Era

Many business leaders are now applying serious pressure towards their IT counterparts to move to a cloud model sooner than later.

Their motivations are unambiguous. Business people see cloud models delivering better IT services at a lower overall cost.

And no one wants to forego a significant competitive advantage.

But it can be harder than it looks -- at least, given many of the familiar public cloud options in the market.

Most larger enterprise IT landscapes are deeply integrated; almost woven together.

Applications aren't usually isolated; they feed, and are fed by, others. Critical business processes that power any enterprise can span dozens of individual application components. And like a central nervous system, the enterprise IT control plane spans all of it, keeping a watchful eye on performance and security.

Untangling the components incrementally, and attempting to move them to a public cloud model one at a time, is turning out to be far harder than it might look to be on vendor powerpoint.

Unfortunately, the basic nature of popular public clouds isn't genetically compatible with what enterprise IT is doing today. And therein lies a thorny problem.

What to do?

It's All Connected -- Or Should Be

I keep arguing with folks that enterprise IT is distinctly different than other forms of IT.  

What a web-scale company wants from IT and what a manufacturing company wants are usually two different things.

One of the defining characteristics of most enterprise IT is deep integration: whether intentional or not.

Data flows are integrated: transactional systems feed reporting systems feed analytical systems which often make transactional decisions -- like what offers you'll be presented on a web page.

Control planes are typically integrated around business processes, or should be: any weak link in the chain can disrupt the desired business outcome.

IT skills and expertise transcends neat boundaries; IT pros do what's needed to ensure that data continues to flow.

Now, Try And Separate Out One Component

When considering a move to a public cloud model, the discussion usually drifts to "which applications might put in a public cloud?".

But if applications tend to be woven together, separating one or two components out for a re-implementation can be much harder than it looks.

The underlying technology of many public clouds is fundamentally different than what you'd find in the data center.

It's managed, secured and operated completely differently.

Existing applications can't be run unless they're tested, qualified and -- in many cases -- entirely rewritten.

And, oh yes, there's this network latency thing :)

If, as an IT leader, you've been successful in creating a deeply integrated on-premises environment, congratulations.  You did your job.

Unfortunately, the act of doing so has made your ability to deconstruct individual application components and move them incrementally to a public cloud model just that much harder.

I remember many years ago when my wife and I foolishly attempted to move house over a period of many days instead of all at once.  I don't know what our reason was at the time, all I can recall is that it didn't work out so well.

The moving boxes would arrive a few at a time, but we had little control as to what was in them.

We'd try to cook a meal, but couldn't come up with enough utensils. We'd try to sleep and shower in the new place, but ended up missing pillows. I tried to fix a few things around the house, but my tools hadn't arrived yet.

It's getting dark outside, where are our lamps?

It wasn't until most of our belongings had arrived that we could get back to a normal life.

Charting A Course

So, what to do?

On one hand, there's no denying the acute business interest in moving to a cloud model. On the other hand, there's no denying the stupefying complexity and deep integration of many IT environments.

The classic rubric of trying to change a jet engine while in flight applies here.

One school of thought is to embark on a massive initiative to re-engineer IT to be more "cloud native" in anticipation of an eventual move to a public cloud model.  

That's a laudable and notable goal -- but, in some sense, we've been continually re-engineering our environments over the last few decades.

Certainly, any investment along those lines (for example, starting to use containers to package applications) is a positive thing. 

My belief? Depending on a complete re-engineering of a significant component of enterprise IT isn't going to save the day. 

The required level of investment -- in time, money and resources -- can only be measured in many years. It's an inherently risky undertaking.  

And the business people don't appear to be willing to wait -- or spend the inordinate amounts required to accelerate to a reasonable timeframe.

An Alternative Approach

If most public clouds don't work the way enterprise IT works, how about a public cloud that's better designed to work with enterprise IT as it is today; and not at some distant future state?

The change-the-world let's-start-over-again visionary IT types will reject most forms of compromise when discussing cloud.  

Unfortunately, most enterprise IT is defined by compromise.

Let's take workloads and application models.  Many IT shops have a wide inventory: older monolithic legacy applications, current multi-tier setups, and maybe a few modern containerized models they're working towards.  

Ideally, your choice of public cloud vendor should support your choice of application and computing model.  Just like you do in the data center. Want to run your own stack?  You should be able to.

How about database?  The majority of valuable enterprise information lives in some form of a database.  

It's not an unreasonable view that any choice of public cloud ought to be able to run databases extremely well -- at least as well as they run in data centers today.

Which explains why Larry Ellison used most of his keynote at Oracle Open World to clearly contrast the ridiculously extreme differences between Oracle Cloud and AWS.

How about being able to have identical environments in the public cloud and on-premises?  Being able to bring the public cloud model behind your firewall?  Have workload portability across both the data center and multiple public clouds?  Being able to manage and secure it all with a consistent control plane?

These are not unreasonable expectations from an enterprise IT point of view.

Why Public Cloud Adoption Has Been Slow For Enterprise IT

When you look at the industry statistics for public cloud adoption patterns by enterprise IT, you'll notice they are characteristically low.  

Dig deeper into actual usage, and you'll find what I call "cloud at the edge" and typically not at the core of the enterprise.  

Maybe desktop and collaboration has been moved to a public cloud model.  Maybe there's some SaaS being done by parts of the business.  Maybe there's a familiar private cloud consolidating generic workloads in the back room.

What's left is the hard stuff: enterprise applications, woven together, powering the critical business processes that every organization depends on to get work done.

Maybe the problem with enterprise IT adoption isn't about enterprise IT.  It is what it is.

Maybe it's more about the poor public cloud choices enterprise IT has been offered up to this point.

-----------------------

Like this post?  Why not subscribe via email?

Comments

[Eric Singer]

I've been trying to say this to my buddy (MS Azure pre-sales architect) for year and he doesn't get it (or ignores it). What I've found even worse, is that C level IT leadership just doesn't seem to get it either.

That said, I'm not a fan of Oracle (in any way) so I'm not cheering for your cloud solution. Sadly, I'm sure Oracle would be a great solution, but they've tarnished themselves as an untrustworthy / mafia like IT vendor. Its hard to imagine an IT vendor being hated worse than MS, but Oracle found a way to do it in spades. It doesn't matter what blog I go to, or who I speak with, Oracle hate spews forth with reckless abandon. Even if Oracle were to turn around their whole philosophy, it will take at least a decade before people are willing to forgive and reevaluate them. In that time, I could easily see AWS or MS being able to accommodate enterprise IT. Its really not a matter of "can they" rather do they want to.

Despite that, the concepts you've put forth lately about the lack of enterprise understanding in the public cloud are spot on.

[Chuck Hollis]

Thanks for your comments, Eric.

When it comes to large, successful IT vendors, you can always find supporters and detractors. That's reality.

I think the smart IT people look beyond the casual perceptions, understand what these vendors bring to the table, and how that's different than other alternatives.

The reason I came to Oracle was simple: I believed they had a unique approach to solving a huge enterprise IT challenge: how do I get to a cloud model without starting over?

Thanks again

-- Chuck

[Michael Endrizzi]

I was standing on a chair cheering for you like a rock star singing his hit single. This is the crux of the private/hybrid/public cloud migration and only a few people get it. You are a digital poet.

"My App Is Slow" -- How would you think AWS/Azure/etc would answer the phone to this question? At least when the enterprise controls the infrastructure there is a management structure in place to force techies to untangle the spaghetti.

[Dan Fisher]

Well said Chuck. I especially liked the quote "Untangling the components incrementally, and attempting to move them to a public cloud model one at a time, is turning out to be far harder than it might look to be on vendor PowerPoint." How true it is! It is interesting to note that the Oracle Cloud is unique in that regard. When placed side-by-side, clearly one of these Clouds is not like the others. Still, it's Public Cloud, and that scares (and confuses) a lot of people.

So, if Public Cloud is not the best fit, there are few other choices left. The No Cloud option (Infrastructure as Usual) and the Private Cloud option (Infrastructure as a Service - but you have to build the service yourself). I know you're not a fan of the Private Cloud model but I still see it as a viable alternative to Public Cloud offerings. A variation of the Private Cloud may be what Gartner calls 'Cloud Inspired' architectures? Perhaps not a true cloud but still allows for infrastructure choice with some automation/orchestration thrown in making a few things simpler? This is probably where most shops will end up as they wrestle with cloud, shrinking budgets, and demands for more transparency and flexibility. Oracle has a few offerings in this space too, so I think they are just hedging their bets a little on the fickle imaginations of Enterprise IT.