As a result of Kubernetes pushing cloud environments to all-time highs, cloud expenditure has skyrocketed, with over half (49%) of respondents to CNCF's most recent microsurvey report focusing on Cloud Native and Kubernetes FinOps. Survey participants said that the implementation of the widely used orchestration platform resulted in a moderate or major rise in their expenses. Less than one-third (28%) reported no change in expenses.
Across the board, Kubernetes is eating varying amounts of cloud expenditures, according to Cloud Native Computing Foundation’s report. The majority, or half of the group, reported using Kubernetes for up to 25% of their budget; however, a respectable portion reported using even more. About 28% of respondents said Kubernetes was using up to half their budget; for 10%, it was up to 75%; and for the other 5%, it was anything from 76% to 100%.
The largest group, including 26% of respondents, spends up to $50,000 per month on cloud computing, while the second-largest group, comprising 22% of respondents, spends 20 times that amount, or more than $1 million per month, more than any other group. For the third largest category, the percentage decreased considerably, with 21% spending less than $10,000 per month on the cloud.
Kubernetes infrastructure comes in different sizes, with a long tail at the bigger end of the spectrum and a sweet spot at the lower end. The majority, 49%, have up to 50 nodes, followed by 17% with 101–250 nodes and 15% with 51–100 nodes, which makes up the third biggest category. The estates on the long tail, which total up to over 1,000, represent 19% of the responses.
The rise in expenditure as well as unintended and unforeseen expenses in cloud settings was attributed to several technological and human causes. By a wide margin, over-provisioning ranked top (70%), followed by a lack of understanding or accountability (45%) at the individual or team level. Third-place finishes were from failing to deactivate resources after usage and having technical debt (43% each).
Overprovisioning, spam, technical debt, lack of awareness or responsibility at the individual or team level, resource-hungry workloads, varying consumption demands, inadequate planning and forecasting of cloud consumption, absence of centralized, consistent, or standardized processes and/or tools for insight and action across all cloud providers, and availability of a self-service infrastructure are some of the factors that contribute to overspending.
AWS Cost Explorer, GCP Cost Tools, Azure Cost Management, Kubecost
The research examined the usage of multiple tools and methods, such as billing analysis, monitoring, dashboards with alerts, and Kubernetes, to comprehend and control cloud cost. The most popular tool, according to the statistics, was AWS Cost Explorer (used 55% of the time), followed by GCP Cost Tools (28%), Azure Cost Management (23%), Microsoft, and independent cost monitoring company Kubecost.
The most well-liked project among respondents was OpenCost, a vendor-neutral CNCF Sandbox project for tracking and distributing real-time cloud infrastructure and container expenses. It provides billing API connections with AWS, Azure, and GCP. For the first time, OpenCost was included among the top 30 CNCF project velocity list, indicating that it has been growing steadily in tandem with the global FinOps movement.
Most respondents said that raising teams' and individuals' awareness of costs and their responsibility for spending would aid in taking control. At 58%, better teamwork and communication around spending and consumption ranked second, while best practices that could be used individually and in groups came in third. Adopting standardized tools and techniques together with senior leadership, according to half of the respondents, would be beneficial.
98% of respondents said it was essential that the engineering, product, and development teams keep an eye on expenses and participate in cost-cutting measures, and 75% said they could convince these teams to do so. The survey respondents represented a range of adoption phases for FinOps techniques and technologies. Thirteen percent have begun to optimize their infrastructure by tapping underutilized instances and rightsizing, while thirty-five percent are investigating and assessing tools for budget analysis and forecasting. Twenty percent have shifted to monitoring, reporting, and reviewing consumption and expenditure.
This microsurvey was developed by CNCF with support from the OpenCost team. It was conducted from June to November 2023 and received little over 100 responses. The whole survey dataset is available on GitHub.
