Cloud sticker shock: all too common, but somewhat avoidable

Inefficient cloud operations are emerging as a top barrier to realizing the full value of cloud.
Written by Joe McKendrick, Contributing Writer
Photo: Joe McKendrick

Cloud's original value proposition was that it would save companies money by sparing them upfront investments in hardware and software. However, it turns out a lot of money is being spent on cloud services anyway, and perhaps even going to waste. That's the gist of two survey reports that track cloud spending.

One estimate is that enterprises are missing out on $24 billion in wasted or misapplied cloud spending. 451 Research's Market Monitor service values the global Infrastructure-as-a-Service market at $48 billion for 2021. More than a third of respondents to the survey, 36%, report they are consuming cloud only at on demand rates -- the most expensive option. "That's a shame, because our Cloud Price Index finds that, across the five hyperscalers, average savings of 36% on the cost of a simple application -- consisting of virtual machines, storage and networking -- can be made just by using commitment discounts," the study's authors state.

A recent survey of 350 IT decision makers by Virtana blames disjointed point tools, silos, lack of visibility, unexpected costs, lack of programmatic optimization. A majority (82%) of organizations with workloads running in public clouds have incurred "unnecessary" cloud costs, the survey's authors state.  

Inefficient cloud operations are emerging as a top barrier to realizing the full value of cloud. For example, 68% of all respondents stated their teams operate in silos, and 70% of respondents said limited collaboration hinders their ability to adapt quickly and improve business outcomes. 

The reasons for cloud sticker shock include the following:

  • Workloads bursting above agreed capacity                                         41%
  • Overprovisioning of compute or storage resources                             35%
  • Storage blocks that are no longer attached to a compute instance     34%
  • Poor job scheduling                                                                              22%
  • Over-buying or having unused reserved instances                              18%

The survey also shows a lack of visibility across hybrid and multi-cloud environments: 84% of respondents are running workloads in multiple public clouds yet 86% of respondents said they cannot get a global view of cloud costs within minutes, creating delays and potentially reducing agility. 71% of respondents agreed that limited visibility across the hybrid cloud environment hinders their ability to maximize value, creates inefficiencies, and wastes time.  

In addition, 66% of respondents stated it is hard to understand if they are delivering the service levels the business needs, and 65% agreed that when there is an issue, they are hard-pressed to identify the business impact. In addition, 77% cited increased performance issues as one of the reasons that pressure on cloud teams continues to rise.  

In addition, 72% of respondents said they are "fed up" piecing together disparate management tools to monitor and manage everything from infrastructure performance to cloud cost and migration readiness. Another 62% report they cobble together multiple tools, systems, and custom scripts to get a global view of cloud costs.  

"A bit of time and effort can deliver huge savings, and cloud providers already make tools available to do this," according to Owen Rogers and Jean Atelsek, the S&P/451 researchers. "For the providers, the benefits are better cash flow, greater predictability and lower costs. Many third-party tools can also optimize cloud use, even across multiple clouds, and building an application that spans venues can yield vast savings on direct cloud costs. However, this isn't easy, and companies face a raft of technical, process and people challenges in doing so. The first step for all cloud users should be to look at what they do today and see if optimization can work for them."

The S&P/451 authors point out "that many enterprises that default to on-demand consumption just haven't considered how significant the savings can be or invested in the time and services to explore them. Others may feel their requirements are too 'bursty' to optimize. On-demand provides huge flexibility, but the reality is that most enterprises don't necessarily need to scale up and down on a second-by-second basis. A balanced approach is to use commitment discounts for long-term baseline capacity, then supplement with on-demand as needed." 

There are three basic techniques for optimizing cloud spending, the S&P/451 report states: 

  • Commitment discounts "offer savings of 70% or more in exchange for making an up-front purchase or committing to a set level of monthly spending."
  • Rightsizing "exploits cloud's inherent flexibility to better match resources consumed with workload demand. This approach is best suited to workloads with unpredictable or variable demand."
  • Cost arbitrage "takes advantage of differences in compute pricing -- either due to idle capacity at a given hyperscaler datacenter or regional variations -- to dynamically tune an application. This method lends itself to long-running workloads." 
Editorial standards