Making the Consumption-Based Cloud Model Work

A consumption-based approach to the cloud — whether measured by bandwidth, CPU cycles or another metric — makes sense for organizations that are small, just starting out or engaged in highly variable workloads. However, to optimize the cloud environment, particularly regarding cost, there are important considerations to be aware of. 

A major benefit of the consumption model is that it accelerates time to market. Developers can spin resources up and down quickly, without the roadblocks of procuring and configuring assets. That said, choosing the right model isn’t an either-or decision, and cost isn’t always the primary driver. 

Let’s look at some of the potential pitfalls of the purely consumption-based model and how to avoid them.

The consumption-based model is a great fit for rapid development and prototyping. Certain types of workloads, however, such as artificial intelligence, machine learning and Internet of Things applications, may have long operational runtimes. At a certain point, the consumption-based model may outweigh the expense of owning your own gear. 

Organizations may have moved to this model initially for reasons other than saving money — often, to increase agility and scalability. However, in some situations, cost becomes an important factor. It’s critical to look at workloads holistically to ensure the desired aims are being achieved.

“Orphaned” resources are another issue to watch out for. It’s easy to spin up resources, but behind the scenes, all the interconnecting components get spun up automatically too: network interfaces, IP addresses, disks, storage and so on. When you get rid of a service or application, some of those resources become orphaned, and they are still within your tenant. 

For organizations with a large cloud footprint, this can be a significant problem. Most likely, no one is reviewing the long list of objects populated in the cloud to identify these orphans. One of the downfalls to consumption-based modeling is that you pay for what you have, and sometimes what you have isn’t linked to anything.

In other cases, the organization may hold on to resources on purpose — for instance, keeping data that could be needed for an audit. If you don’t move that to a less expensive cloud tier, you may be paying much more than you need to.

The cloud often is more economical for businesses that are just starting out. However, many organizations reach a point in growth where the consumption-based model becomes a negative, like an anchor weighing them down. That’s why so many organizations are moving to a hybrid model. They may also lower costs by using a baseline of reserved instances, which they supplement with on-demand resources and, potentially, on-premises architecture.

Figuring out the right equation — overall and for specific workloads — isn’t easy, but it’s a necessary step to control costs and optimize the environment. In fact, helping organizations define their deployment models is one of the most important services we provide for our customers. The consumption-based model can be powerful, but the best way to leverage it is by having a strong understanding of the business case for doing so.

Story by Brent Blawat, a multicloud principal cloud solution architect for CDW. He has written four books on Microsoft PowerShell. 

Urban Haas, an executive technology strategist for CDW. He has been with the company for more than 17 years and specializes in enabling business/digital transformation for his customers using cloud technology.