Most people I talk to don’t realize that there actually is an official definition of cloud computing or “the cloud.” The National Institute of Standards and Technology has created a definition that runs to several hundred words and essentially defines cloud as, “compute resources, delivered over a network.”
While this is an accurate definition, it is a little broad, and it doesn’t say much about the value of cloud. However, there are two components of NIST’s definition that are a little more specific, and can be used to illustrate the value of cloud:
- The essential characteristics of cloud
- Service models of cloud
Private infrastructure vs. Cloud: The indisputable difference and value
Both of the components found in the NIST definition, noted above, contain elements of unique differentiation and value that only cloud can deliver. For this posting, let’s consider the 5 Essential Characteristics of cloud; specifically, let’s consider them in the context of private infrastructure.
This is a perspective that CDW uniquely maintains from our position of selling and deploying both private infrastructure and public cloud solutions. There are very few private infrastructure deployments that have even one of these characteristics, because cloud capabilities are not easy for IT departments to replicate using private infrastructure. However, there are no private infrastructure deployments that can replicate all of these characteristics. And most importantly, the characteristics that cannot be replicated have a real value to pretty much every organization out there.
Let’s consider each of the 5 characteristics:
- Can private infrastructure provide on-demand, self-service? With the right software for virtualization, automation and orchestration, this can be done.
- Can private infrastructure provide measured service? Again, given the right software and integrations, this can be accomplished with private infrastructure. Charge back and show back, by the hour or by the month, for example.
- Can private infrastructure provide broad network access? Private infrastructure can be made accessible from anywhere and can be located anywhere (at least if you include colocation in your data center strategy). Not to say it is trivial to achieve the kind of broad network access that is inherent in the large public cloud compute offerings like Azure, but technically, the answer here is still, yes.
Well, so far so good, but how about those last two characteristics?
- Can private infrastructure provide resource pooling and elasticity? No!
Resource pooling means that multiple companies are served by a large pool of resources. Private infrastructure serves only the company that owns it. Elasticity means that resources can not only be scaled up, but also down. Private infrastructure does not allow for either of these. Furthermore, there is tremendous value in these two characteristics which only public cloud can deliver. I’ll give two examples to illustrate that value.
- High Performance Computing (HPC). Not long ago a world record was set with public cloud. In about an hour, 71,000 cores were spun up in a cloud-based HPC cluster to crunch through an engineering design workload that normally would have taken a month using an expensive “to buy” and expensive “to maintain” in-house HPC cluster. In the cloud, that same workload ran in eight hours and the infrastructure cost was about $5,500. The black-and-white cost savings are only part of the story here. HPC in the cloud means that a once-per-month workload can now be run in hours. This turns the design process into an iterative process that can be run quickly and cheaply over and over until the perfect design is uncovered.
- Disaster Recovery as a Service (DRaaS). With private infrastructure, if you want a DR solution, you find yourself in the “two houses” situation. That is the situation where you have to build a second house, fully furnish it, and maintain it just so you have a place to live temporarily while your first house is rebuilt after a fire. Nobody does that. They rent for a few months while their one house is rebuilt. Yet with private infrastructure, that is exactly what you have to do. In contrast, the resource pooling and rapid elasticity of cloud computing allows you to “rent” your DR when you need it (or when you want to test it).
These are just two examples of the way cloud computing can provide value that private infrastructure will never be able to replicate. If you think about your company, can you find other examples? Perhaps you have a peak season, or an annual event or some other need for compute resources that is not 24x7x365. The reality is that most organizations do have this sort of requirement. This is why I say that organizations that do not include cloud computing in their overall enterprise architecture are competing with one hand tied behind their back. Your competitors are probably looking at ways to cut costs or improve processes through the unique capabilities of cloud computing. Shouldn’t you as well?
And, of course, who doesn’t want the on-demand, self-service, metered, pay-as-you-grow and broad network access characteristics found in cloud computing. Those can be achieved with both public and private cloud, and also have real value to your organization. Whether you pursue a public-only cloud architecture, or a hybrid cloud architecture, the value of the cloud computing model is real, and now is the time to begin to incorporate it as part of your enterprise architecture.