Despite the fact that cloud is part of the daily conversation in many enterprises, I still find a significant gap in many places in terms of a true understanding of what it means. This is somewhat compounded by the reliance on standard definitions of cloud computing from NIST and other sources. These definitions are helpful in some respects, but they are far more focused on attributes than on business value – and the business value is what is truly needed in the enterprise to break through the barriers to cloud computing.
First, let’s divide the enterprise IT landscape into three buckets:
- Infrastructure & Operations: this is the core group in IT responsible for operating the data centers, running the servers, keeping the lights on, ensuring adequate capacity, performing IT governance, etc. We’ll call them I&O.
- Applications: whether custom application development by a dev team, or a COTS application licensed in, or a SaaS app running externally, the applications are where the core value of IT is created. As a general rule, I&O exist to serve applications – not the other way around (though I think we can all come up with situations in our past where the nature of that relationship has not been so clear).
- The Business: these are the users and application owners. Developers build apps for “the business” and I&O hosts them. Often times, especially in large enterprises, application development actually sits within the business under a business line CIO. So the app owners also control the application development and are the “customers” of I&O.
When talking about cloud, it’s really critical to have this context in mind. If you are talking to the business, they care about some very specific things related to their applications, and they have requirements that I&O needs to address. If you are talking to I&O, they have a related but very different set of issues and requirements and you need to address them in terms that are relevant to them. Let’s start with the business.
Talking Cloud to the Business
If you are speaking with the application owners within a business, they care about the following (generally unsatisfied) requirements with respect to their infrastructure and IT needs (the ordering is not important – it will be different for different businesses):
- Control – The pre-cloud world is one in which the business makes requests of I&O, often through an onerous and labor intensive service request workflow involving forms, approvals, emails, negotiation, rework, and more. The business puts up with this because they have to, or they just go outside and procure what they need from external vendors. As with many innovations, cloud computing first entered through the business and only later got adopted by IT. As a business, I really want to be able to control my IT consumption at a granular level, know that it will get delivered reliably and quickly with no errors, etc. This is the concept of “on-demand self-service.” Let me configure my requirements online, push a button, and get it exactly as I ordered with no fuss.
- Transparency – I heard a story once where a company had hired so many IT finance analysts that there were more people accounting for IT than actually producing it. It may be a myth, but I can see how it might actually happen. If I apply the management accounting principles of the shop floor to IT, I start to get into activity-based costing, very complex allocation formulas. But even with that, it’s still viewed by the business as more of a black box than transparent. I sat with some IT folks from Massachusetts a couple years ago and they all groused about how costs were allocated – with the exception of one guy at the table who knew he was getting a good deal. Winners and losers, right? What the business wants today is transparency. Let me know the cost per unit of IT, in advance, and give me control (see 1) over how much I consume and let me know what I’ve used and the costs incurred along the way. No surprises, no guess work, no hassle. In the NIST cloud world we call this “measured” IT.
- Productivity & Innovation – Pre-cloud I&O processes are often so onerous that they significantly impact developer productivity. If it takes me several meetings, days of analysis, and hours of paperwork to properly size, configure and formulate a request for a VM, that’s a huge productivity drain. Further, if I have to wait several days or even weeks before this VM is available to me, and I have to wait for it, that slows me down. At one financial institution I spoke with the VM request form as 4 packed pages long, required 12 approval steps, and each approval step had an SLA of 3 weeks. Yes, that’s a potential of 36 weeks to return a VM and still hit their SLAs to the business. In reality it never took 36 weeks – but it often took 6-10 weeks for a VM. Seriously, why can I just have a VM now, when I need it? That’s what the business wants. Related to productivity, innovation is seriously stifled in most enterprise IT environments. Imagine if I’m on a roll, have this great idea, but need a VM to test it. Then imagine a series of hurdles – sizing, configuration, paperwork, approvals and waiting!! Now, it may be a pretty cool idea, but unless it’s part of my top priority task list and was approved months ago, it just isn’t going to happen. The business wants support for innovation too. That means it wants speed. This is the concept of “elasticity” in IT. Give me as much as I want/need now, and when I’m done, you can have it back.
- Cost – Last but often not least, the business wants a smaller bill from IT – and the benchmark is no longer in their peer group. The benchmark is Amazon, Google, Microsoft, Rackspace and others. The benchmark is the cloud. Why pay $800/month for a virtual machine when Rackspace will rent it to me for $100? Not only does the business want better IT – more control, transparency, productivity, and innovation – but they also want it at a lower cost. Easy right?
When engaging with the business application owners about their cloud needs (you do this, right?), and they are having a hard time articulating what is important to them and why they want cloud, ask them if they want more control, transparency, productivity & innovation, and lower cost. If they don’t really want most of this, then perhaps they don’t want or need a cloud.
Talking Cloud to IT Infrastructure & Operations (I&O)
In short, I&O really would like to satisfy the requirements of the business listed above. Remember that I&O’s mission is to serve the business by serving their applications. When talking with the I&O side of the house (make no mistake, there are at least 2 sides here), talk in terms of the requirements of the business. Yup – control, transparency, productivity & innovation, and cost.
How? Be a cloud provider to the business, of course. But what does that mean? So many people I meet still think that a self-service portal in front of a vSphere cluster is all that it means to be a cloud. It’s more than this – it’s a completely end-to-end automated operations model to deliver IT services. In order to meet all of the requirements above, including at a reasonable cost, everything that can be automated should be automated. So-called “enterprise clouds” that still require manual steps in the provisioning process cannot achieve the cost advantages of a fully automated environment (unless of course the cost of putting in the automation, divided by the number of units produced, is greater than the cost of doing it manually). This is no different than with the making of products. Even in many heavily automated mass-production businesses such as auto manufacturing, IT still done in a way where every VM and deployment landscape is an exception crafted by hand. That’s a huge waste!
Cloud computing operating models (cloud not a technology, it’s the application of many technologies to change the way IT is operated) grew out of necessity. How could Google, Amazon or other large scale web business possibly handle tens of thousands of servers without either armies of low cost workers or “extreme” automation? They could not, and neither can you even if your data center footprint is in the hundreds of server range. Clearly the automation route is less expensive in the long run, at least for the vast majority of tasks and actions that are performed in a data center every day.
Now enterprise IT gets to have many of the same techniques used by cloud providers applied in their own operations. With all of the software out there for building infrastructure (IaaS) and platform (PaaS) clouds, it’s never been easy to envision and implement the “IT Factory of the Future” in any sized environment. Take an OpenStack, BMC Cloud Lifecycle Management, VMware vCloud or other cloud stack and create your infrastructure factory. Then add Apprenda, Cloud Foundry, or one of dozens of other PaaS frameworks and and create your application platform factory If fully implemented and integrated, the variable labor cost for new units of IT production (VM, scaled front end, etc.) will approach zero.
Let’s take this to an extreme. Some in IT have titles like VP or director of infrastructure. They run the I&O function. Let’s give them a new title – IT plant manager if they run one data center. Or VP of IT production if they run all I&O. Even if you don’t go that route, perhaps that’s how these people need to see themselves going forward.