Cloud Economics and the Six Most Damaging Mistakes to Avoid
The value potential of cloud is enormous, but only for companies that understand and adapt to the realities of cloud economics.
Companies might be moving to the cloud, but their thinking is stuck in the legacy world of on-premises computing. That thinking has proven hard to change for many companies, with economic and financial models grounded in decades of traditional IT practices that are based on “owning” IT instead of “consuming” it.
As a consequence, companies are developing business cases, negotiating contracts, and making economic calculations that don’t take into account the different financial approaches and models that are specific to cloud. Not only is this resulting in value derived from the cloud falling far short of expectations, but it is also, in some cases, threatening cloud programs themselves, with some businesses even considering reversing course.
Among the many cloud-economics mistakes companies make, the most persistent are:
- Making a business case that conflates the economics of day one and year one
When making a business case for moving to cloud, accurate estimates of cloud value are complicated by a focus on the “lift and shift” approach—that is, on a targeted migration of existing applications with limited remediation.
This approach allows enterprises to quickly develop a cloud footprint and start building cloud capabilities on day one. The economic benefits come mainly from reduced hosting, storage, and maintenance costs. Unfortunately, those benefits are often muted because companies retain most of the technical debt and operational inefficiencies of those migrated applications, which keeps them from taking advantage of dynamic infrastructure provisioning made possible by cloud.
- Using ‘average cost’ capital-expenditure economics
Traditional IT operates under a capital-expenditure model, where enterprises engage in episodic, long-range demand-planning exercises, followed by capital outlays and ongoing depreciation. In this model, data-center capacity is built out years into the future, the marginal cost of consuming additional infrastructure capacity is minimal, and companies measure their cost efficiency by looking at their average cost and infrastructure-utilization level.
By making it possible to dynamically add near-limitless capacity, cloud service providers (CSPs) have changed the paradigm to an operating-expenditure model, where enterprises pay for what they consume. As a result, the most efficient cloud economics now hinge on the ability to effectively evaluate capacity demand—and the corresponding incremental or marginal costs—at any given moment. In essence, this is about paying for capacity only when you need it, rather than paying for capacity you don’t use. Companies instead need to develop a dynamic operating-expenditure approach to cloud economics that continuously optimizes incremental costs by choosing the cloud services that best match their current workload requirements.
For example, one media company dynamically scales up its compute capacity ahead of major customer promotions to accommodate increased user traffic and scales it down after the promotion ends to avoid unnecessary cloud spend.
- Forecasting cloud spend based on historical factors only
As organizations make the leap from the capital-expenditure world of traditional IT to the operating-expenditure world of cloud, history becomes a much less reliable predictor of the future. This becomes a big issue when companies need to estimate cloud spend to develop budgets or make allocations to support new cloud-based products. While companies make allowances based on cloud’s prevailing operating-expenditure model, old habits are hard to break, and forecasting typically still relies heavily on the capital-expenditure model. This often results in a greater than 20 percent discrepancy between forecast and actual spend, leading to poor allocation decisions and arduous rebudgeting.
The key to better forecasting and budget planning for cloud is to tie it more closely to business priorities. For example, if a company is planning a large promotion tied to Black Friday, it is likely to see a surge in customer interest. Similarly, plans to shift pricing to a subscription model will lead to new consumer behaviors. Since cloud costs vary by usage, these kinds of business decisions will have an impact on them.
- Automatically extending the elasticity benefits of compute to other cloud services
The elasticity and scalability of cloud is economically ideal for workloads with variable cloud-consumption patterns. A video-streaming enterprise was able to establish a unit-cost relationship between the cost of cloud-computing services and the corresponding business demand drivers (such as compute cost per subscriber) based on statistical analysis. This allowed the company to match its compute needs to its business demand patterns and predict cloud consumption with more than 95 percent accuracy. This ability to accurately match demand with need allowed the company to better allocate spend.
- Divorcing the cloud-economics road map from the cloud-architecture road map
When building the cloud business case, businesses often assume optimistic cloud-utilization levels. This inflates projected savings because, despite the promise of dynamically scalable cloud capacity that can be tailored to match application demand, the reality is that most companies end up with lower cloud-resource utilization than they’d hoped for. While some enterprises with advanced cloud-native architecture see resource utilization rates greater than 60 percent, most companies fall below 30 percent—and, in some cases, below 10 percent.
High utilization rates are at least partly dependent on an architecture capable of supporting them. For example, autoscaling of compute resources can significantly improve utilization, but only if the application architecture is upgraded. Unfortunately, the business’s cloud-economics and ‑architecture road maps are often developed in relative isolation from each other, leading to business cases focused on utilization rates that cannot be supported. For this reason, companies need to tightly link the cloud business case with the cloud-architecture transformation.
- Migrating all workloads to cloud, no matter their scale or type
The economies of scale have allowed hyperscalers to deliver better returns, in the form of cost savings and/or better business outcomes, than what many companies can do by themselves on-premises.
That doesn’t mean, however, that every workload should be migrated to the cloud. The recent cases of companies repatriating major workloads, especially storage services, from cloud to their own custom-designed on-premises infrastructure are a case in point. The scale and homogeneity of these workloads may create on-premises economics that are equivalent to or better than those offered by cloud providers. For this reason, companies that have an environment with a small number of massively scaled workloads need to be selective about adopting cloud.
The cloud is a rapidly evolving space that demands close attention to shifts in financial modeling. For businesses to capture the promised value, they need a strong FinOps capability to make sound business decisions and manage consumption continuously based on a fundamental understanding of cloud economics. As cloud consumption increases exponentially and becomes ever more core to the business, the ability to effectively manage cloud economics will differentiate companies that have cloud aspirations from those that have found cloud value.
Text source: https://www.mckinsey.com/
Image source: https://www.computerweekly.com/