Clouds are exceptionally difficult to discuss in the abstract.
They become much more real and tangible when you associate them with a particular business challenge.
As a result, I've started to collect various "cloud meets business" success stories. They're all very cool, but they take some serious time to write up.
I thought I'd start with an internal one first: how EMC's vLab cloud is changing an important aspect of the way EMC does business.
Although there are aspects that are unique to EMC and our particular situation, I believe there are many useful lessons here that can be more broadly applied.
The Expanded View Of Cloud
So many people evaluating cloud concepts tend to think in rather limiting terms of better ways to do exactly what you're doing today. The more expanded view is when cloud enables you to do things that weren't possible in the first place.
Point that expanded view at a handful of critical business processes, and magic can happen.
Such is the case with EMC's vLab.
Stepping back a bit, one of EMC's critical business processes is -- simply put -- selling a lot of stuff.
Since many of our products are technically rich, a key part of the sales cycle is putting our stuff in front of potential buyers so they can evaluate our products and offerings, and compare them to others. There's only so much you can get from static documentation; at some point you've got to put your hands on something to really understand it.
It's fair to say that "getting cool stuff in front of customers" is a big deal for EMC. The better we get at it, the more we move the needle.
The Starting Point
Since EMC has been selling technology for a long time, most of the legacy processes were designed for the physical world. You either had to bring the technology to the customer, or bring the customer to the technology.
The historical approach involved shipping a bunch of equipment somewhere, assembling it all, and running some exercises -- then disassembling it all and sending it back if there wasn't an immediate purchase decision. Multiply that basic motion by hundreds or thousands of times, and it doesn't take a rocket scientist to figure out that's painfully inefficient for everyone involved: for us as well as our customers.
Although I'm sure the trucking companies made a lot of money :)
The second approach involved the customer either traveling to a site with the equipment, or perhaps logging into an environment that was created specifically for the purpose at hand. Both have their inherent limitations: effort to hand-craft the environment, use of dedicated vs. shared resources, etc. etc. Better -- but not by much.
At EMC, Matt Coviello -- as part of the EMC Solution Group -- has been persistently chipping away at coming up with a better way of enabling this key business process, using cloud concepts. He's now on his second cloud, and I found his experiences very indicative of what happens when you point a very powerful solution at a very big opportunity.
At a high-level, vLab is a self-service environment for demonstrating and evaluating EMC software stacks. If it runs in a virtual machine, it can be supported by vLab. Please keep in mind that many ostensibly physical EMC products (like storage!) are now available in VSA format, meaning that a great deal of functionality can be exercised via virtual machines.
Matt's team can now offer up a growing menu of 100+ preconfigured and on-demand "experiences" based around various EMC and partner technologies. Individual experiences (or groups of experiences) can be provisioned in 15 minutes or less.
Today, he can flex up to 5000 concurrent virtual machines, and is bringing on capacity to reach 15,000 concurrent experiences by early next year. The complete team -- all in -- is 7 people and 1 manager in various global locations.
User experiences are semi-persistent: entire evaluation environments can be created, stored, retrieved, etc. The underlying architecture and management processes are Vblock-based, basically the same standards used by the broader EMC IT organization.
I was able to chat with Matt, and he had some great insights to share.
I mentioned before that Matt's team was on their second cloud. I think it's fair to say that their first cloud was more of the homegrown variety, using the various bits and pieces that were available at the time. Matt is very clear that his second Vblock-based cloud is worlds better than his first approach.
First, performance, user experience and efficiency stats are worlds better. Matt's limited team can now spend more time delivering user experiences that people want vs. integrating, managing and supporting the plumbing.
Since the vLab cloud is nothing more than a different consumption option on top of the standardized platform EMC is using corporate-wide, he gets all sorts of benefits: technical assists when needed, the ability to flex workloads against a larger (and compatible!) pool, and so on.
He does admit that there was a school of thought early on that the requirements for vLab were somehow "different" enough to justify a separate technology stack, operational model, etc.
He would be the first to say "get over it". Building on a standardized platform has helped his team focus far more on value delivery vs. infrastructure operations.
The Enablement Role
Matt's vLab organization is essentially an enablement function for other business units. His team isn't chartered with directly reaching the end consumers of the service he's delivering -- that's up to the part of the business that wants to deliver an experience for an audience.
The first -- and largest -- internal customer is EMC's global presales organization. They provide input on the services to be delivered, policies associated with service delivery, managing (and paying for!) consumption, and so on.
He exposes a service catalog, they consume.
Matt adjusts the services offered based on what his internal customer wants, and so on -- just like any external IT service provider would do.
A special case exists in EMC's vSpecialist team. Thanks to Chad's efforts, you'll see the EMC vLab powering the EMC-oriented hands-on lab at both VMworlds and EMC World, for example. Matt's team can flex up a considerable amount of capacity for those events, and then flex back afterwards.
The other part of EMC that's starting to appreciate vLab capabilities are EMC's product groups. They're starting to realize that the easier it is for customers to get their hands on various EMC technologies, the more likely it is that they'll want to purchase that. Some internal groups (e.g. the Atmos team) are all over this in a big way, others are just starting to dip their toes in the water. It's not a question of policy per se, it's more a matter of the level of effort required to deliver a useful technology experience using vLab.
Finally, EMC Education is another big user. Using new technology often requires new skills, and the ability to put a real, live (albeit virtualized) environment in front of students -- wherever in the world they might be -- is a huge enabler.
The Inside-Out Path
If you're a regular reader of this blog, you'll likely remember that I've often mentioned that process maturity is one of the big gating factors to increased use of cloud deployments. When it comes to vLab, you'll see the same principle in play.
The first vLab services were offered only to a small group of authorized corporate users. The second batch was aimed at the EMC presales organization. The third was around offering services in structured settings, such as VMworld or an educational offering. The next target audiences are EMC partners, to be followed by EMC customers using a self-service model, and eventually -- anyone at all who has an interest in test-driving EMC products.
At each step, the team learns, adjusts and improves. As processes are understood and validated, the next step is taken.
The Capacity Challenge
Make something easy to consume, and people will consume a lot more. The EMC vLab model makes IT resources incredibly easy to consume, which -- inevitably -- leads to incredible consumption. Matt's vLab consumption statistics are off the charts, and he's only begun to tap into the potential.
Which brings up the interesting challenge -- how is all of this paid for?
At an aggregate level, it's pretty easy to see the benefit to EMC (and our customers!) so no one is arguing that funding vLab expansion is a bad idea. However, once you get beyond that lofty goal, you dive into the real world of fixed budgets vs. variable consumption.
Current IT financial practices enforce a fixed budget mindset: here's how much you get for the year, now go make it happen. But, in the cloud world, IT is variably consumed. In just about all cases, you *want* to drive big spikes in consumption: that's generally business goodness.
The closest analogy to this problem exists with EMC's component suppliers and the associated supply chain. They (and we!) would like to be able to lock into well-known volumes and capacities determined far in advance -- and then negotiate around that.
But, in the real world, the market has this way of moving here and there in ways we can't easily predict.
Get it wrong, and there's either a lot of unsold inventory or a bunch of paying customers who will start looking elsewhere. Providing variable IT services really isn't all that much different when you think about it.
Look more closely at the physical supply-chain processes, and you'll see a great degree of "flex" built in (and paid for) around hedging different scenarios. All participants in the supply chain have constant communications to constantly adjust the parameters. This sort of constant communication is rather unusual for most IT situations, but is turning out to be a hallmark of many SP models.
The challenge will be to bring this same sort of model to demand and supply forecasting for variable IT services: here's our best guess, and here are the scenarios we'll invoke if we get it wrong. No easy answer here.
Automate The Future, Not The Past
There seems to be a lot of people out there shopping for magic automation tools; usually well ahead of any sort of practical experience around what they'll need in the new environment.
Matt's team continually introduces new forms of automation based on visible requirements: here's an important workflow, here's where automation can help, what are our options?
That's a sharp departure from the more typical holy grail search for the Perfect Automation Tool.
The Road Ahead For vLab
Like any EMC function, Matt's team isn't resting on their laurels -- there's a very long list of things they really want to get to. First and foremost, there's the need to expose vLab-like experiences to larger and larger audiences, e.g. partners, customers and prospects. Second, he's doing what he can to expose vLab capabilities to other EMC functions -- even though he's not 100% sure where all the funding might eventually come from :)
More interestingly, his team is starting to work with more complex, multi-VSA scenarios -- ones where you might have dozens -- or potentially hundreds -- of individual VMs working together in a coordinated fashion. As more EMC-owned Vblock capacity gets deployed around the globe by EMC IT, he wants to be an anchor tenant that can flex workloads closer to users (latency is ugly) in addition to the present capabilities at the new EMC Cloud COE (Center of Excellence), based in Durham, North Carolina.
The View Ahead?
If you're like me, you don't spend any time anymore trying to define or explain cloud. It's sort of like trying to define or explain the internet -- people either get, or they don't. Instead, I've spent the last year or so trying to explain how IT leaders introduce cloud concepts into more traditional environments -- and why cloud changes how IT does its business in the process.
I think the next year will be about success stories -- how progressive IT teams are applying cloud concepts to big business challenges and end up really moving the needle. The stories are starting to emerge quickly, and they're all interesting.
I'm just glad my own company is one of them :)