The Weather Company bases its forecasts on input from a network of more than a billion sensors – some of them weather stations owned by the company, some of them owned by the government, and some of them privately run by volunteers who contribute the data they collect. This makes The Weather Company the most accurate weather forecaster in the world –it’s right about 75 per cent of the time. That’s a pretty forgiving margin for being the best in your field. Weather’s like that. Forecasts change.
The computing cloud in your forecast is likely to change, too. Saying that your organization is moving to cloud computing is like saying there will be weather. It’s accurate, but not extremely useful. Years ago, in an effort to get traction among consumers for the cloud model, companies equated “cloud” and “Internet.” People threw the cloud tag at anything that computed. That was neither accurate nor useful.
Fortunately, over the years, the brighter bulbs in the industry have managed to winnow what was once more than 30 (Insert here)-as-a-service models to a workable definition of the cloud. While characteristics make the cloud—elasticity, resource pooling, on-demand self-service—the definitions that are useful to planning your cloud infrastructure fall into two categories: service models and delivery models.
Service models
Delivery models—which we’ll get to in a moment—get most of the attention in any cloud discussion. But let’s put the horse in front of the cart. How your cloud services are delivered is immaterial if you haven’t chosen the service model that best suits your organization. Your best service model for today won’t necessarily be your best service model for tomorrow, so evaluate periodically how your company is growing and where your company is going.
The good folks at the National Institute of Standards and Technology (NIST) in the U.S. have defined three service models that cover the waterfront that all those X-as-a-service models were cluttering. Software-as-a-service (SaaS) is the model that’s most familiar to most people, because we use it, whether we recognize it or not, in our personal lives as well as our corporate lives—e-mail services, data storage services, calendaring services. In the enterprise world, this extends to customer relationship management (CRM) services, enterprise resource planning (ERP) services, accounting services, and so on. SaaS is the lowest-touch service model. Essentially, you pay a subscription fee for whatever software it is you need, you login to use it, and your service provider handles everything else.
SaaS is a good model for a newly launched organization or an SMB that’s outgrowing its standalone PC-based infrastructure. SaaS is the model that most shifts IT costs from capital expenses (CAPEX) to operating expenses (OPEX). Since capital can be hard to come by for a new company, and growing SMBs often have other capital priorities to keep up with, shuffling IT expenses to the OPEX side of the ledger is an attractive proposition—it’s cash-flow-based, not equity-based.
At the other end of the cloud service scale is Infrastructure-as-a-service (IaaS); the service provider supplies fundamental computing, networking and storage resources, while the customer provides everything from the operating system up. This option better suits an established organization that is, for example, looking at a hardware refresh window. Rather than investing in expensive new hardware, and the inevitable maintenance costs that come with it, the company can achieve some cost certainty by moving some IT expenses to the OPEX side. Unlike SaaS users, IaaS customers will still have to retain enough IT staff to run the business almost as if its IT infrastructure was on-premise.
In between is Platform-as-a-service (PaaS). The service provider supplies hardware, OS, libraries and other tools for the customer to deploy its proprietary applications. This is a good option for organizations that have invested considerable resources in their own application development and want to divest themselves of infrastructure and maintenance costs, while still retaining control of their applications stacks.
None of these three models are mutually exclusive, and different applications may be better suited for different service models—CRM might work better on a SaaS model, while data warehousing might be better suited to an IaaS model. You’re allowed to mix and match – no customers will use just one model.
Deployment models
The delivery model serves the service model, and while NIST defines four—public, private, community and hybrid—I’m going to go out on a limb and say only the last will be relevant 10 years from now. The differentiator between public (and community) and private cloud delivery models is shared infrastructure, not, as is often believed, location. A private cloud is not necessarily an on-premise cloud (or, as I like to call it, a “data centre”). A private cloud is devoted to a single organization, while retaining the scalability, resource pooling, and self-service characteristics of cloud computing. In a public cloud environment, you’re sharing hardware with other companies, possibly even competitors. The public cloud model offers pricing considerations, while the private cloud model offers control.
A hybrid model is simply a combination of two of these deployment models. Like with service delivery models, different deployment models suit different applications. Sensitive customer records might be better suited to a customer-devoted, or even line-of-business devoted, private cloud, while seasonal spikes in transactions might better be served with the scalability of a public cloud model. The two working in tandem is the future, especially as we bring Internet of Things data capture scale and real-time analytics into the picture.
So, if there’s cloud in your forecast, you’ve got considerable planning on your plate – something trusted service providers like Rogers are ready to help you with. And remember that forecasts change; today’s cloud might not be the one you need for tomorrow.