Like many companies, we at EMC start our new year with a leadership gathering. We gather to celebrate, connect, strategize and share. They are *always* great events.
I found this year's gathering was particularly rewarding in terms of deep content. The majority of the meeting was spent unpacking the depth behind the core elements of EMC's strategy: cloud, big data and trust.
We dove in from a product and technology perspective. We came at it from a services view. Another take from a services and skills viewpoint. And, finally, the organizational and business model implications.
For me, it was like a wonderful meal that just went on and on. Rich, detailed and exceptionally well-thought out -- although your head started to hurt after a while.
Underlying much of the discussion was the central notion of a software-defined datacenter (SDDC for short), representing the next generation of infrastructure and operational models. All through the discussion, that was clearly the conceptual foundation for so much of what needed to happen in the industry.
And I started to realize we still have a lot of explaining to do: not only around the concepts themselves, but what they mean to IT groups and the organizations they support.
I've now had some time to think and digest, and I wanted to add a few different perspectives to the mix.
We're all familiar with the "blind men with an elephant" parable: each describes what he's perceiving very differently, but it's very hard to figure out there's an elephant in the room.
Any time a new concept comes on the scene, we get to see this story play out again: we all start out by seeing new things through our traditional lenses.
The moral of the story is simple: there is no "correct" view -- at least, until you integrate all the individual perspectives.
By now, many of us have had practical experience in explaining SDDC concepts to various audiences and stakeholders. Like most early-stage concepts, it's not one-size-fits-all when it comes to 'splaining things.
Here's the current collection I've been using.
View #1 -- A Natural Extension Of Server Virtualization
When Paul Maritz first became CEO of VMware several years ago, he said something at an EMC leadership meeting that still sticks with me to this day: virtualization -- like the microprocessor -- is one of those infinitely extensible technologies: it can be applied in place after place -- and not just compute servers.
The core principles are deadly simple.
Abstract -- so you're working with logical representations of things vs. the things themselves. Virtual servers vs. physical servers, to name a familiar example. A measure of efficiency and agility results, but there's more to come.
Pool -- now that entities are abstracted, they're far easier to pool and share. More efficiency and agility inevitably results.
Automate -- now that your resources are abstracted and pooled, they are far easier to orchestrate and automate. And you get another dollop of efficiency and agility as a result.
We've all seen how dramatically transformative those simple concepts have been in the server world. There are now many more workloads running in virtual servers than physical ones.
Although it's early days, we're seeing the same phenomenon starting to play out in the network world. And it's even earlier days in the storage world.
But those are just physical entities we're talking about. The concepts can be further applied to logical entities: databases, security software, app servers, middleware, data fabrics, mobile apps, and so on.
As Paul said -- infinitely extensible.
And, every time you abstract/pool/automate, you get the double payoff of efficiency and agility.
Do enough of it, in enough places, in a consistent and standardized fashion, and you're now starting to approach the notion of a software-defined datacenter.
View #2 -- A Natural Evolution Of Server Virtualization
The server virtualization revolution continues to burn through the IT landscape. Thankfully, it's no longer about whether virtualizing servers is a good thing or bad thing; it's more about how and when you're going to do it. And the resulting efficiency/agility payoffs are now considered the "new norm" in IT thinking.
There's no going back now.
But so many of the resource entities that virtualized servers interact with are not virtualized, or -- at least -- not virtualized to the same extent that we see in the server world.
That's a problem. IT architecture and service delivery is a team sport -- and some of the other team members have some work to do.
If you look at the efficiency/agility payoff we've been able to collectively achieve with server virtualization, you start thinking hard about what you need to do to get the rest of the team playing at the same level: network, storage, database, security, apps, frameworks, etc.
View #3 -- A Philosophical Statement on "Good" IT Infrastructure
One of the things I love about IT is that it's a continual battleground of ideas and philosophies -- with no referee!
In the infrastructure space, one of the arguments bubbling up is what represents "good" IT infrastructure: purpose-built, or software-defined? In some ways, it's reminiscent of the old ASIC-vs-microprocessor debate -- purpose-built, or software-defined?
One one side (perhaps well-represented today by Oracle and their Exa-products) is the view that "best" results can be achieved by engineering all the hardware components to work together: server, storage, network, database, etc.
This is not a new view: consider IBM's mainframe and iSeries, DEC VAX and perhaps Teradata.
Their argument largely hinges on what defines "best". If one values agility and responsiveness to new demands, any hard-engineered approach will necessarily suffer -- and there are plenty of examples of this everywhere.
If one is after more traditional IT aspects: e.g. performance, availability, efficiency, etc. it's still debatable -- thanks to the incredibly amazing pace of evolution we see with industry-standard technologies.
Taking a larger view, is the "best" enterprise model a collection of various purpose-built hard-wired stacks, with integration and orchestration more of an afterthought? Or is the "best" model one where functionality is dynamically defined by software through virtualization?
A software-defined datacenter takes a purist view: variable IT services are best delivered by abstracted, pooled and automated entities. Performance, availability, efficiency, etc. is something you dial in dynamically -- and not with a soldering iron.
Personal prediction: by 2014 this discussion will be largely over, and we hopefully can all move on :)
View #4 -- Supporting The Next-Gen IT Operational Model
If you've followed me at all over the last few years, you know I've had much to say about the wave of IT organizational transformation that's clearly at hand.
At the risk of boring my regular readers, it's pretty simple: IT fashions itself as the competitive IT service provider of choice. These transformed IT groups measure themselves by delivering the services that the business wants to consume -- just like everyone else in the business world.
Make no mistake, it's a big deal: new roles, new skills, strong leadership, etc. -- it's a thorough revamping of the traditional enterprise IT operational model.
But once this transformation is well underway, the inevitable happens: the modern IT team starts going out and looking for modern tools to do their job.
Through this new lens, anything that isn't completely abstract-able, pool-able and automate-able just isn't all that interesting in their world.
Server virtualization is clearly there -- but what about everything else?
In my mind, it is this growing class of transformed IT functions that will be the first and best customers of SDDC technologies.
View #5 -- Supporting The New IT Mission
Does IT exist to save money -- or to make money?
In the past, it was clearly more of the latter -- but in today's digital economy it's hard to argue that IT doesn't have a new mission: to create the "digital factory" that produces almost all of the new value propositions going forward.
Inevitably, that means a river of new IT capabilities, delivered on new platforms, using new devices -- and in entirely new ways. In business, speed is king. The more competitive your marketplace, the more agility matters. And if a business leader can't get agility from the IT team, they will inevitably be forced to go outside to someone who can.
Quick story: in one of my customer meetings, I was sharing how our EMC IT team migrated to a new SAP instance on our private cloud this summer. One of the charts stated that we were now running at ~10% CPU utilization, and one individual loudly commented about how wasteful that seemed. Old world meets new reality.
I shared that -- before -- when we were running at over 90%+ CPU utilization, we really couldn't do anything useful or interesting with our back-end systems. But, now that we had all sorts of headroom, we could consider any number of cool things that would be extremely valuable to our business, and back-end horsepower wouldn't be a concern.
It was an investment in the future, not the past. Most certainly, a shift in perspective.
View #6 -- A New "IT Stack" Model For The Industry
Much has been said already about the "stack wars" in the IT industry. You see it everywhere: IT vendors lining up various pieces to present as complete (and as integrated) a solution as possible for their customers.
Done right, I see it as a good thing: customers can choose to go with a vendor's stack, or pick and choose as circumstances dictate. More options are good.
But if you believe the IT world is going completely virtual -- and there seems to be no shortage of IT shops who see it that way -- the strong preference will be for interoperable stacks that are as fully virtualized as possible.
Over time, the same industry dynamics we've seen play out in the physical IT world should play out in the fully virtualized world: vendors working hard to offer complete and integrated virtualized stacks.
And -- done right -- customers will have the choice of going with a single vendor's stack, or picking and choosing as circumstances dictate.
View #7 -- Yet Another Forcing Function For IT Transformation?
When the server virtualization wave started to get fully underway, you predictably met two kinds of IT groups: those who were well on their journey (and clearly seeing the benefits), and those who were taking a more wait-and-see incremental approach.
No surprise, I saw that the first group were clearly performing better at delivering more responsive and efficient IT services to their stakeholders. And, for the second group, it was sort of business-as-usual. Lots of roadblocks, lots of process, lots of skepticism, etc.
But I clearly saw an "organizational performance" gap forming between the haves and the have-nots. You'd go back and visit the latter group a year or so later, and -- very often -- something big had changed in their world: a new leadership team, a new mission -- or maybe they were now being outsourced.
In the business world, no one wants to be playing at a disadvantage for very long, and that includes the IT function.
In one sense, the adoption of server virtualization was a rough proxy for best-in-class IT infrastructure and service delivery. And IT had to learn to organize for success around the new way of doing things.
As more SDDC technologies find their way into the market, I think this is going to happen all over again. The resulting efficiency / agility payoff will be compelling. There will be early adopters, and the wait-and-see crowd. The organizational performance gap between the two will inevitably widen, with the predictable consequences.
As Mark Twain once said: "history doesn't repeat itself, but it does rhyme".
Is There An Elephant In The Room?
Maybe there's no elephant in the room today, but there's clearly one on the way, and sooner than you might think.
You have to wonder -- how long will it take us to collectively figure out there's an elephant in the room, and we ought to be doing something about it?
Preferably before it gets here?
I would humbly offer that taking off those dark sunglasses might be a good first step :)