In an effort to elevate and enhance the dialog around Cloud Computing, we recently drafted anumber of blog articleshoping to better clarify the nature of and proper definition forCloud. Ironically, I found myself in an engaging discussion with our Director of Cloud Services just this week, a dialog that bordered on debate at several points over the finer points of Cloud definition. As I better understood the point he was making, it occurred to me that I myself had completely failed to adequately address this aspect of the cloud compute model in my past posts.
The core issue we were hashing through concerns physical vs. logical architecture as it pertains Private, Hybrid and Public Clouds. The market largely accepts that Private Cloud implies a pool of dedicated infrastructure sitting beyond some type of DMZ, whether on premise or in a hosted environment. To transform this into a Hybrid Cloud, you simply extend access through a secure, virtual tunnel to another distinct cloud resource (either a Community or Public Cloud), creating a composite architecture for bursting or resiliency. Granted, this requires some fairly complex technology and orchestration to enable seamless data and application portability and synchronization, but that’s a different matter. Our Director was challenging the basic notion that Private Cloud’s must be deployed on dedicated infrastructure, contending instead that the way users access a Cloud resource is must more central in how it is classified as type of cloud resource. In other words, if a Cloud is accessible via a public network connection, then it is fundamentally a Public Cloud, where as a cloud resource that is limited in accessibility via VPN connection constitutes a Private Cloud. And in turn, whether the underlying Cloud infrastructure is dedicated or multi-tenant is essentially irrelevant in terms of its classification.
I decided some further research was in order, so I went to theNIST Special Publication 800-145, which defines Cloud Computing as “a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.” But that really doesn’t help answer the question. Continuing through SP 800-145, it defines the Private Cloud as a “cloud infrastructure provisioned for exclusive use by a single organization comprising multiple consumers (e.g., business units) that may be owned, managed, and operated by the organization, a third party, or some combination of them, and it may exist on or off premises.” Again, this doesn’t really address the conception of dedication vs. multi-tenancy and logical separation. I went to another recognized authority (Wikipedia), and found they were pulling their definitions straight from NIST. But I do note that they include a simple illustration for 3 basic types of Cloud Model, and that clearly seems to support the designation that Private Cloud = Dedicated Infrastructure.
This really was making me curious, so I did a fair bit of additional research, which included exploration on the sites of various Cloud Vendors themselves. It is interesting to note a fairly common use of the term ‘Dedicated’ in product packaging these days, which I suspect is largely born out of confusion by their customers and prospects over this exact issue. And in the end, the larger issue here surrounds helping drive a common taxonomy so we can mitigate confusion as we engage with others in a Cloud dialog. I know I will be sure to clarify with customers and prospects in the future when discussing Private Cloud deployment whether their cloud needs to be deployed on dedicated infrastructure, and how they intend to expose access to this resource for user’s outside the corporate firewall.