Last week I ended up in a Twitter chat on the subject of private cloud. One of the others on the thread stated that private cloud was a myth. I responded, but as Twitter only allows 140 characters, I felt it was important for me to come back to the subject and describe my point of view in more detail.
The argument can be summarized as this: Private cloud is a myth as many people call their virtual environment "cloud." As a result private cloud is not OSSM (On-Demand, Scalable, Self-provisioned, Measured). For that reason, it is a myth. It does not exist. Let me add to that... A number of vendors consciously mix virtualization and cloud in their marketing material to make it sound as if it is enough to buy their products to build a private cloud.
Yes, many people call virtual environments private clouds, but it’s not because people make a mistake in the naming of a particular service that the real service does not exist. So, not that many companies may have implemented a private cloud in their datacenter, but some have. They have combined virtualization with automation and self-provisioning to provide their users with an on-demand environment that is scalable (within the limits of the available resources) and self-provisioned. So, as far as I can see, this is a cloud.
Let’s go back to the definition of cloud computing. And let’s take a definition that is well accepted by a large majority across the globe, the definition of NIST, the National Institute of Standards and Technologies. The final version of this definition was released last October and looks like this:
“cloud computing is 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.”
Reading through that definition, I do not see anything that excludes the private cloud. You could argue that the end speaks about service provider interaction, and so separates the service provider and the user.
OK, I accept that, but will point out the trend that the IT department becomes a service provider to the business. Since the appearance of the “shared services center” concept, the idea of an IT department that operates as a service provider to the business users has become increasingly popular.
Cloud Computing takes the concept one step further, transforming IT in a “service broker,” a department that provides the users with services from multiple sources, traditional environments, their own cloud environments and external suppliers.
With that in mind, the private cloud has its reasons of existence. It is the on-site portion of the IT environment delivering services to the end-user. It will host the “core” services of the enterprise. And here I use the term in the sense provided by Geoffrey Moore when he talks about core versus context.
Can we apply the OSSM model described earlier? Let’s start with the “self-provisioned” part. HP’s converged cloud vision calls for a single user experience. In other words this implies a single portal and user catalog giving access to all accessible services, regardless of where they are sourced from. The private cloud is probably the best place to host this functionality, serving as the enterprise gateway to services provided by the traditional environment and private, managed and public clouds. It’s a self-service portal which the user accesses to provision/de-provision and use services. In that portal, the user can also get information about the services he has provisioned, how much he has consumed and what service level agreements he actually received.
Let’s now move to the “on-demand” part. But before we do that, it’s probably good to review what “on-demand” actually means. So, let me quote the SeachDataCenter definition:
“On-demand (OD) computing is an increasingly popular enterprise model in which computing resources are made available to the user as needed. The resources may be maintained within the user's enterprise, or made available by a service provider.”
If through self-provisioning I can initiate a service, well, it seems I am in an “on-demand” environment. The combination of virtualization and automation allow me to provision that service transparently. Again, within the concept of converged cloud, whether the actual resource is provided within the enterprise (through a private cloud) or made available by a service provider (public cloud) does not make a difference. So, I believe we can check that one off too.
Let’s now move to the “scalable” part. Most enterprises have IT resources within their datacenters. These include servers, storage and networking. Many companies have started virtualizing these environments, putting an abstraction layer in between the hardware (often heterogeneous) and the actual environment used by the applications.
Combining this with automation allows the scale up and down of the resources consumed by the applications according to demand. This part is mandatory as otherwise you do not have a cloud. So, I agree that virtualization is not cloud, but again, that does not mean a private cloud cannot exist.
Now, in a private cloud you are limited by the actual infrastructure you own. Actually in any cloud you are limited by the actual infrastructure available, only in a public cloud the resource pool tends to be much larger.
But what converged cloud brings is the possibility to use the resource pool of the private cloud as long as resources are available and then the provisioning of public cloud resources when the private ones run out. Policy rules can even define which services should stay in the private cloud and which can be migrated to the public one.
So, do you agree, we can check this one off also?
This leaves us with the final characteristics “measured.” This one is a little trickier. Not all enterprises measure the consumption of their users. This used to be done in mainframes, but as IT evolved, and IT capacity became plenty, companies have lost the practice. In traditional environments, consumption is rarely measured. The evolution to shared service centers did bring the subject back on the table, and most cloud environments available today have some form of measurement. They may measure the actual consumption of infrastructure, the time between the provisioning and de-provisioning of a service by a user, etc. What is done with that information is another subject altogether. Is it used to charge departments, to spread out IT costs, or is it just used for informational purpose? That depends on the company. The point is, most private cloud environments have the functionality to provide IT with consumption information.
Again, within the frame of converged cloud this becomes very important as you want to be able to charge the user for the services he/she consumes regardless of where those services are sourced from. That obviously includes private cloud. So, I believe we can consider private cloud to address the “measured” characteristic.
Yes some are pretending that virtual environments are private cloud, and by doing so depict private cloud incorrectly. In the same vein, some are depicting lions as wild cats. That does not mean cats are a myth, it just means people are not correctly depicting things.
I believe the same is true for private cloud. Private cloud is NOT a myth. Private cloud environments exist, the technology is available today, and the private cloud has a fundamental role in a converged cloud strategy. As companies increasingly move to hybrid environments, they will implement true private clouds on their premises or have them managed by service providers. These environments will be dedicated to them (private), and serve as the point of entry of their Converged Cloud.