Do ‘private’ or ‘internal’ clouds deserve to be called “clouds”? Prominent cloud thought leaders including Sam Johnston and George Reese think not. Some say, to truly be considered “cloud”, a solution must be outsourced, use virtualization, and billed by the hours used, like a power utility.
If those who hold this view are right, then it’s impossible for an internal private cloud to be “cloud”. I disagree. In fact, the most disruptive, game-changing events in the rise of cloud computing may emerge with internal clouds behind the corporate firewall.
The Argument Against Internal Clouds
The basic argument against “internal clouds” being true clouds goes like this:
Consequently, an internal “cloud” doesn’t qualify when the organization owns and manages the underlying infrastructure, because it doesn’t achieve the fundamental mission of outsourcing IT pain. This argument however is flawed. Whether cloud services are outsourced or run internally, so long as the cloud consumer is shielded from the “pain” of slow and expensive service, they receive the benefits of the cloud. The argument against this assumes outsourcing of the IT management pain is of central importance to the entire business when in truth, the cloud consumer, with whom the majority of the benefits lay, couldn’t care less.
Is who owns and operates the underlying hardware relevant question? What if the ‘internal cloud’ exists in an internal datacenter, which is operated by an external partner, like IBM for example? Is that a ‘cloud’ or not? What if it’s part of a unit that is spun off into a separate business and a new contractual relationship is created? Does that suddenly make it a ‘cloud’ when it wasn’t before? Clearly the nuances of modern business relationships and dynamics make even the notion of ‘outsourcing’ very fluid and hard to pin down.
While the idea that “outsourcing makes it cloud” is wrong, it is true that a key component of “cloud” is “it’s not my problem”. That is a far more interesting discussion.
Self-Service Inside the Firewall
A better way of saying “it’s not my problem” is ‘self-service‘. Self-service is about reducing or eliminating friction. The consumer of cloud services wants to consume computing services quickly and efficiently, without the associated pain. They don’t care whether the cloud is internal or external.
This is what “it’s not my problem” is about and why enterprise IT can be successful in delivering a consumer-centric, low friction experience to cloud consumers if they so choose. The question, in my mind, is more of whether enterprise IT will be able to shift their current slow-moving processes and procedures in such a way that they can compete effectively inside the firewall against external cloud services.
If Salesforce.com’s success is any indicator, it’s going to be a tough battle.
Getting in the Fight: The Argument for Internal Clouds
IT departments are facing the issue of how best to remain competitive with external services like Amazon’s EC2 (public infrastructure clouds) and Salesforce.com (SaaS) offerings. Savvy CIOs and IT executives want to enter the game and compete effectively. Enterprise vendors are gearing up to deliver offerings to allow IT departments to deliver competitive internal cloud services. This includes both incumbents and startups: VMware’s vCloud, Open Nebula, Cisco UCS, Mezeo, Nexenta, EMC, and more. CIOs have many options to build internal cloud offerings and they will do what it takes to stay competitive.
The challenge, however, is that none of these technology offerings necessarily empower the IT organization to change, become more agile, and to increase self-service capabilities. More important than any self-service portal is understanding how to transform your IT organization.
(re)Defining Cloud
So where is there agreement on what constitutes a true ‘cloud’? James Urquhart’s position is that cloud is “just an operational model” and I generally concur. It’s important to remember that there are two views to consider when defining cloud: the consumer’s and the cloud operator’s.
To the consumer, any cloud service must be low friction and self-service. The billing model doesn’t matter. How and when you pay is a friction point that changes depending on who you are. So to be ‘cloud’ in this context means providing payment flexibility for the consumer at hand: utility pay-by-the-hour model, pay one year in advance, use a PO, use a credit card, or whatever, so long as it is convenient for the consumer. They need to be up and running as fast as possible, with the option to leave any time, without having to put up with human interference or friction of any kind.
To the operator, clouds must be highly automated with minimal overhead. How else can they deliver the self-service that cloud consumers desire? If you need to pick up the phone to call a sales person to provision your cloud service, the operator has failed. If you need to spend a week reading the documentation, the operator has failed. As on-demand services, clouds must be highly automated and easy-to-use.
This is why we typically define cloud simply as:
Self-service IT capabilities delivered via
automation
You will notice that this definition encompasses all areas of cloud, is simple to understand, and is a useful litmus test. Ask yourself: do the two pioneers in the cloud computing space, Google and Amazon, both provide self-service IT using large amounts of automation? What about the canonical SaaS play, Salesforce? Yes, yes, and yes.
Conclusion
Cloud computing is not about ‘outsourcing’. Instead, it’s about delivering self-service IT ‘at scale’ through automation. Only the largest businesses should be building and maintaining their own infrastructure. If you are a large enterprise, you can compete successfully with external cloud services from Google and Amazon, but ultimately, your ability to succeed will be judged on your service delivery capabilities.
Competing means changing the way you do IT today including transforming the team and culture. You will ultimately determine which applications need to be outsourced and it will be important to help your organization choose the right ones based on end user requirements, corporate security policy, and governance issues. For everything that remains inside the corporate firewall, building and delivering it using an internal cloud is not only a valid direction, it’s really the only direction. If not delivered in a cost-effective and agile manner using internal clouds, even those mission critical applications may be outsourced.