Perhaps there is no deeper disappointment in life than when a cherished concept fails to produce the desired results. Such is the case with the industry’s notion of private clouds.
I’m throwing in the towel, walking away – and cursing under my breath. It’s a failed concept.
Yes, I was seriously seduced early on. Check out this Jan 2009 post where I breathlessly made the case for this new private cloud model. If I wasn't the first person to do so, I was certainly close to being the first.
I argued vigorously for the cause for many years. For me, it was the right answer at the right time.
I hereby publicly admit the error of my ways. The world has changed, and so must I.
Why is it that traditional private clouds have left most IT shops at a dead end?
The idea behind a private cloud was straightforward: use virtualization and a little automation to create an easy-to-consume data center infrastructure service that provides virtual machines and their associated storage.
By keeping a little extra capacity around, IT could be much more responsive to that next provisioning request.
At the time, private clouds were far more efficient as compared to manually provisioning virtual resources, or – worse – physical resources. And, make no mistake, they were a big improvement over what came before.
And then the world changed.
The Business Wants An Enterprise Cloud
If private clouds were the supply side, what’s happened to the demand side?
For so many business models, the digital economy is here and now: disrupt or be disrupted. The business now craves the speed, agility, functionality – and cost structure -- of public clouds.
Competitors new and old are arming themselves with the new weaponry. For many, it’s a matter of urgency.
Against modern demands, traditional private clouds now look like an antiquated response.
Here’s why.
1 -- The Private Cloud Consumption Model Is All Wrong
Why? It ain't cloud.
IT buying racks of equipment and attempting to offer a “service” (just like the cloud guys!) has turned out to be misguided at best. IT is still in the capital expenditure business – they have to guess how much stuff they’re going to need.
Buy too much and you waste resources. Buy too little and you’ll come up short at the wrong time. Nobody ever gets it right twice.
We’re back to “have a hunch, provision a bunch”. IT is asking the business to spend ahead of actual demand, and bear the risk for IT’s inability to accurately forecast. Public cloud vendors don’t force this on their customers.
The consumption model is wrong on another level. Business consumers crave cost transparency: show me how much it actually costs for what I’m consuming, and no one else. The reason is simple: business people need to understand true costs.
It is the rare IT organization that has achieved this cost transparency back to its business users. Why? It’s very hard to do. Equipment, licenses and labor is bought in bulk, but consumed as a service. And the required skills to price variably consumed services from those static ingredients isn’t trivial.
That’s why you typically find a “flat corporate tax” back to IT consumers. As any armchair economics student will tell you, that’s about the most destructive pricing mechanism one can imagine – it incentivizes all sorts of suboptimal behaviors.
When pressed, IT organizations will often create something that looks like a price list for their services catalog, but everyone knows the truth – it’s usually a made-up corporate fiction.
2 -- The Private Cloud Functionality Model Is All Wrong
The better public clouds aimed at enterprises have rich suites of ready-to-use functionality. Not the case with private clouds. Or many popular public clouds, for that matter. It’s often a build-your-own world for both.
Want to deliver a speedier environment for your application developers? You’ll have to design, integrate and support your own. Want a powerful platform for your analytics users? Same story – significant assembly and required. Want great integrated infrastructure support for your enterprise application portfolio? Sorry, no can do.
Can private clouds deliver out-of-the-box functionality that is immediately usable by the business? Yes, if all you want is a generic VM.
Examples of standalone usable private cloud functionality are hard to come by, but virtual desktops just might qualify.
OK, fine, but … my view is that simply recreating a legacy desktop paradigm using VMs isn’t exactly a game changer for most business types. The same is true for virtualizing, say, your email server or collab site.
Put differently: private clouds do OK at consolidating existing stuff. They're aren't designed - out of the box -- for the new stuff people want to do. And that ain't cloud.
Yes, IT folks care about consolidating stuff; but business people – well, not so much.
Just ask them.
Although there is no shortage of converged and hyperconverged vendors who are !!hyperexcited!! about consolidating existing workloads. Mea culpa, I’ve been there before. Sometimes I feel they’re offering a better COBOL compiler. Cool stuff, but hard to be relevant in the bigger picture.
Just don’t confuse consolidating existing physical workloads with delivering the new capabilities business users crave.
3 -- The Private Cloud Compatibility Model Is All Wrong
Most private clouds are built around vSphere. Most public clouds aren’t. Most vSphere shops use a vSphere-based set of tools to run their shop. Most public clouds don’t. Most private clouds offer IT a high degree of control. Most public clouds don’t.
Can you see where this is going?
There are strong and powerful reasons why many enterprises will want to use an integrated combination of on-premises and off-premises resources. Each has strong pros and cons. But what happens when you try and use them together?
Can you move simplistic VMs around from here to there? Sure, we’ve all seen that demo, and it never fails to impress the non-enterprise IT crowd.
Let’s try something a bit more practical: how about seamlessly integrating public clouds with on-premises versions of the same, and run demanding enterprise workloads as a single, integrated architecture?
I didn’t think so -- not unless you want to staff a sizable systems integration and support team, and wait around for them to eventually figure out how to glue everything together. That’s not the cloud people want, is it?
It’s simple: most public and private weren’t designed to work together. And that's a problem.
4 -- The Private Cloud Labor Model Is All Wrong
Clouds are supposed to be more efficient: both with resources and labor. Simply shifting resource efficiency into greater labor costs sort of misses the point, no? Many cloud models -- both private and public -- are showing themselves to be labor-intensive in enterprise IT settings.
Why? They don't do what is needed out-of-the box. People are required. Expensive people. The frustrating part? These people are usually re-creating what already exists elsewhere. That's a problem.
Re-inventing an existing proven solution doesn’t create value. Not only that, you’re very likely to do a poor job of it.
When I was working with VCE, I can’t tell you how many infrastructure teams vehemently insisted they could do a better job doing infrastructure system integration than a company dedicated and resourced to productizing infrastructure integration.
I guess it was all they knew how to do. By the way, I never saw them do a better job -- ever.
Now that I’m at Oracle years later, it’s the same story again, but this time around the gap between what my current employer can do with their technology and the in-house integration team’s capabilities is much, much more pronounced. That doesn’t mean they’ve given up, though – no matter how far-fetched their arguments sound.
It just doesn’t create value to do unneeded integration between things that weren’t designed to work together when compatible and pre-integrated alternatives are available.
It’s like opting to design and build a car from piece parts vs. buying one. And arguing that you’re delivering a better car while saving time and money in the process.
Take a look at any enterprise IT org chart, and ask a pointed question: how many people are spending their careers either trying to (a) recreating existing, proven solutions, (b) trying to support incompatible stuff working together, or (c) suffering the consequences of people attempting to do (a) and (b)?
As you’ll see in a moment, it doesn’t have to be that way.
Now, let’s take a business user’s view.
What’s the opportunity cost of IT investing in the ”unneeded systems integration business”? How long do business users have to wait for usable results vs. status meetings? How well does the IT-assembled solution work when eventually delivered? More broadly, why does IT often insist on recreating known and proven solutions, and not doing a good job at it?
Is it to justify their existence?
Not all that long ago, I had to do some personal car stuff: license, registration renewal, etc. I spent the better part of a day driving to a government building, and being shuffled from long line to long line with papers in my hand.
Worse, I saw a well-staffed office of state employees behind the glass partition who didn’t exhibit any signs of urgency.
What frosted me was that I was paying for how they chose to run their business – both with my time and my tax dollars.
It took substantial effort to keep my mouth shut, smile, and get my government-required paperwork done. It was clearly designed to be self-perpetuating as only bureaucracies can be.
Is it any wonder that business people get so frustrated with enterprise IT?
Things Don’t Have To Be That Way
That’s why I consider the hypervisor-centric private cloud a failed model: there’s more to winning at enterprise IT than constructing a VM provisioning machine. It doesn’t even begin to solve the current problem of delivering an enterprise cloud: one that meets the agility and functionality needs of modern enterprise application landscapes.
Time to get off that train, and onto the next one -- one that's going where you need to go.
And, yes, there is certainly a better way. I won’t go into excruciating detail, but allow me a quick example of why things don’t have to be so hard. Yes, this is an Oracle pitch -- but it also illustrates what you could be demanding from your vendors.
Oracle has three parts to their cloud strategy.
1 -- Oracle delivers integrated SaaS, PaaS and IaaS designed for modern enterprises, ready to use today. No assembly required, unless you want something special. It’s already trusted by many thousands of companies around the world. It supports almost any workload, and not just Oracle software.
2 -- Oracle delivers on-premises solutions engineered for modern enterprise applications – database, application logic and middleware, and analytics. All of them come with precise equivalents in the Oracle Cloud. The engineered approach delivers far better results in the data center, and no system integration is required to make private and public work together, as the public cloud is now simply an extension of your data center. If your data center plans change, everything comes with cloud insurance.
3 -- For those craving the agility of the public cloud in the data center, Oracle delivers Cloud Machines – subsets of the Oracle Cloud – that offer the same functionality, economics and operational model of the public cloud. Again, precisely equivalent with the public Oracle Cloud.
Take a moment, please, to digest the impact of all of this, and how it has fundamentally changed my perspective. It should change yours as well.
All at once, my cold disgust with today’s failed private cloud model is gone, now replaced by the warm glow of optimism. There is hope after all. Yes, I do wish we all could have started here, but we didn’t.
To be fair, with this model you’re trusting a single vendor to be held accountable for delivering results. If the vendor doesn’t execute, you clearly have other options. But some find this all-in approach philosophically unattractive.
I would argue that maybe it’s time to get a new religion, as the alternatives are proving strategically disastrous for larger enterprises.
Just consider how this approach neatly solves private cloud shortcoming as outlined above:
For example, the consumption model is instantly fixed, if needed. Consume off-premises, consume on-premises – it’s the same financial model. Mix opex and capex model as needed. If desired, IT can get out of the capex forecasting business, and business users get full unit cost transparency. Voila!
Integrated enterprise functionality is immediately available, because Oracle has already done all the hard work to integrate their products and technologies with others. It also helps that Oracle owns its own technology stack: from storage to database to mobile apps.
As a result, IT can now focus on using modern IT vs. attempting to re-invent it.
There are almost no compatibility issues to “fix” or “engineer” or “integrate” with this model. It’s all the exact same stuff on both sides of the wire, with a single vendor supporting it all.
And – best of all – the labor model gets rightsized and realigned for the cloud era. Yes, there are people who really enjoy screwdrivers and glue guns, but the need for vast armies of them is quickly fading in the new model.
Although – I have to point out -- some of these hands-on integration folks have found a new life, doing the same sorts of gluing together of things in build-your-own clouds such as Amazon’s, essentially recreating their historical world view but now in the cloud. I admire them for their inventiveness.
My view? Unnecessary systems integration didn’t make sense in the physical world, and it certainly doesn’t make sense in the cloud world.
Where Does That Leave Us?
Maybe you like your private cloud. Maybe it delivered the intended benefits. Maybe it moved the needle from what came before, and is seen as successful. That’s all good.
But that was then, and this is now.
I now see traditional private clouds as a strategic and architectural dead end. Before long, you will too.
Like so many technologies before it, the world has moved on, and left private clouds behind.
There is no logical road forward that makes economic or strategic sense. Any proposed potential way forward would inevitably require an army of IT people doing integration and support work of questionable value – and business people patient enough to pay for it, wait for it, and live with its inadequacies.
My view is that more business people are wising up. They may sense that their dear enterprise IT colleagues may have lost their way in the era of cloud.
Surely, there will be more hard questions to come: what are you IT guys doing, how long will it take, how much will it cost – and why are you insisting on doing this the hard way?
If you’re a senior IT leader, best if you have some good answers prepared.
-----------
Enjoy this post? Why not subscribe via email?
From 2012 -- "Why You Really, Truly Don’t Want a Private Cloud" http://organic.ulitzer.com/node/2288203
Posted by: TheEbizWizard | May 24, 2016 at 09:11 PM
Hi Chuck,
I fundamentally disagree.
Yes, I am probably bias (like you are) as we are providing private (but also public) cloud services.
I have seen private cloud implementations fail and succeed. And generally that was not because of the technology. It was about the change or transformation that is required to make a cloud model work.
You have to break down the silo's, change the delivery model to your business, adapt your processes to high level of automation and change the culture of the people delivering the private cloud services.
Yes, technology is not perfect, but as you said, it is a lot better than what IT organizations traditionally can deliver. And btw, because of this I have seen Oracle implementations fail just as much as any other technology in the market.
I agree with your analysis that most public cloud vendors are not compatible on the hypervisor level. Meaning you have to decide how and where to run your workloads. It may mean some kind of migration or reinstallation of your applications.
But with automation this can be managed. I have seen implementation of a VMware based private cloud that through the portal can deploy instances just as easy on AWS as their private vSphere based cloud. Then deploy that same application somewhere else if needed.
That was on an IAAS basis, but that same cloud also ran Cloud Foundry. And applications developed on that platform are even easier to move around or can run distributed over multiple environments.
Sure, most organization are not there yet, we are all on a journey and for some organizations it may make sense to move everything to the public cloud, though I do not see enterprises do this short or even medium term. But they will be moving some workloads to the public cloud.
If you want to be compatible select a cloud vendor that matches your private environment. If it is just for your Cloud Native Apps, run it on Amazon, if it is .net on Azure, etc. based on the requirements of the application.
And if you are an Oracle based organization, check out Oracles solution. But (I am guessing here, but assuming that you run your cloud on OracleVM). Non Oracle workloads probably need some form of migration to run there as well (or does the Oracle cloud also provide a vSphere, KVM or Hyper-V environment).
My take is we are moving to multi cloud environments. The key is how you are going to manage those private and public environments. We started at abstracting server hardware, we are now abstracting storage and network, creating virtual DCs, the next step is the abstraction of clouds. Decouple your application from the environment it runs on. Based on (business) requirements, translated into policies, workloads will be deployed on the environment that provides the best service at the best price. And that will be a mix of traditional (e.g. Oracle) applications all the way to Micro-services running in containers.
I see the first developments in the market. But again. IT IS NOT ABOUT TECHNOLOGY. It is about people, process, organization and culture. And any implementation of cloud (private or public) from any vendor not addressing this is doomed to fail
Posted by: Aernoud van de Graaff | May 25, 2016 at 04:39 AM
Great article, not a total believer but I love how you consider the business and economics aspects.
Your ideas work well with
1) Cookie cutter applications: Netflix
2) Low support requirements: Angrybird games on Facebook
3) Standard or simple security requirements: xkcd comics
4) Applications that have high/low elastic demands: Game of Thrones streaming
5) Predictable cost model
Unfortunately businesses tend to:
1) Want IT to build a 3 legged purpled hair 1 eyed dog
2) Want to call and yell at people when the purple dog pees on their carpet
3) Prefer not to have the NSA or Chinese access to their designs or financial fraud records
4) Well...IT just sucks are building things cheaply no matter if its in the cloud or on-prem. But
once built many applications have a predictable cost model.
Thanks again,
dreez
Posted by: Michael Endrizzi | May 25, 2016 at 09:11 AM
Chuck,
Remember the days when people went screaming from mainframes to their new found flexible low cost freedom from IT ---> IBM PCs?
Seems like the mainframe is baaaaaaack; MainCloud.
dreez
Posted by: Michael Endrizzi | May 25, 2016 at 09:45 AM
Chuck,
Great post. Spot on.
This is relevant with small and large enterprise customers alike. The industry is sitting on the tip of some massive change.
KO
Posted by: Kyle Ohme | June 15, 2016 at 10:14 AM