Kubernetes in the wild

The 2025 edition of the Kubernetes in the Wild report highlights global trends in the technology’s adoption and usage in production environments from thousands of organizations across diverse industries. This report is based on data from organizations using the Dynatrace unified observability and security platform to secure their Kubernetes clusters while ensuring optimal performance, health, and reliability.

Key findings from this past year’s analysis provide insights into the Kubernetes utilization differences between cloud and on-premises environments, particularly regarding cluster hosting and infrastructure model preferences. The report also reveals how organizations are adopting auxiliary workloads, primary areas of growth, and the most popular open-source technologies and programming languages organizations are using in their Kubernetes environments today.

These trends highlight the need for organizations to comprehensively understand their Kubernetes environments.

Currently, two out of every three clusters are hosted in the cloud. In 2022, only 45% were in cloud. This represents a significant and sustained shift toward cloud hosting over recent years. While cloud migration is still a top priority for organizations of all verticals, several organizations have already made major shifts toward the cloud, resulting in a moderation in the adoption rate year over year.

With most clusters now in the cloud, this stabilization marks a new era where cloud-native computing and containerized applications are the standard for modern IT and business operations. In 2025, cloud environments will likely continue to grow at a steady rate and solidify as the “new normal” for delivering cloud-native applications and services.

Furthermore, widespread standardization of Kubernetes lays the groundwork for new and innovative ways to maximize the value of the platform. However, new approaches and best practices will take time to come to fruition and become standardized as organizations implement the necessary Kubernetes-adjacent tools, technologies, and practices to usher in a new wave of innovations.

In past years, nodes with smaller memory sizes, around 16 to 32 GB each, have had a higher concentration in the cloud compared to on-premises. Conversely, on-premises clusters generally contained more nodes with larger memory sizes, around 32 to 64 GB. In 2024, the distribution of node memory sizes does not significantly deviate from this trend, with smaller nodes shifting toward the cloud and larger nodes remaining on-premises.

This indicates the value companies are realizing from the scalability and agility of cloud architectures. Additionally, there were no significant changes in the number of nodes and pods per cluster in both cloud and on-premises environments.

The procurement price of hardware remains only one of many considerations when developing a cost-effective infrastructure model. Scalability and operational efficiency are important factors teams must consider when deciding on node size, quantity, concentration, and location.

For example, while using more, smaller nodes may not lead to lower costs overall, this approach does provide flexibility and cost optimization opportunities for scaling up and down based on demand.

One way Kubernetes demonstrates its “operating system” status is how organizations utilize auxiliary workloads in proportion to application workloads.

Auxiliary workloads support and run alongside the primary application workload. They are used for tasks such as running and supporting system services, monitoring, build pipelines, service meshes, and other services that enhance the core application’s functionality.

The ratio of pods for auxiliary to application workloads shows a steady increase from previous years. This increase suggests organizations are continuing to expand their use of Kubernetes beyond simply deploying, running, and managing applications to also use Kubernetes to standardize how they deploy and manage their auxiliary workloads.

In the past, pods of auxiliary workloads represented roughly two-thirds to one-third application workload and have increased to nearly three-quarters auxiliary to one-quarter application workload. Technologies like service meshes, observability platforms, security controls, and messaging systems play a key role in this evolution, reinforcing Kubernetes’ reputation as the “operating system” of the cloud.

Moreover, examining median pod runtime patterns reveals that pods of build and backing service workloads have much shorter runtimes than pods that directly support applications. While build pipeline pods are short-lived at just 4 minutes, application, monitoring, and system pods run consistently for 24 hours alongside the 8-hour daily runtime of backing service pods.

Let’s look at a median sized cluster and sum up the runtimes of all the running pods for one day. Such a cluster logs approximately 3,500 pod hours a day. At 46%, application workloads account for almost half the pod hours, 37% are system, 14% monitoring, 3% backing, and less than one percent of pod hours are from build workloads.

This distribution of pod lifespans illustrates how critical system and monitoring functions are to maintaining consistent and reliable application functionality. This breakdown underscores the importance of optimizing workload efficiency to ensure cluster stability, scalability, and cost effectiveness.

From just under 35% in 2022, the use of Kubernetes security tools has surged to over 50% in the past year. This upward trend is expected to persist as security awareness increases and both new security threats and solutions continue to emerge.

A new insight from this past year’s analysis shows the rise in adoption of internal development platforms (IDPs). IDPs help teams streamline and automate the provisioning of development environments. Although still in the early stages, the adoption rate surpassed 10% in 2024, emerging as a new category and one that is likely to increase in coming years.

These conditions emphasize the continued need for teams to prioritize and gain insights from security data efficiently by combining observability and security data in a single view. This integration of observability and security enables teams to analyze security data in the context of the larger digital environment, allowing for easier management of both the security and performance of cloud-native applications.

Dynatrace OneAgent can identify the specific programming languages of every application workload running on Kubernetes and provides key insights into programming language preferences and trends across organizations.

Data from OneAgent shows that Java virtual machine (JVM)-based languages remain the most widely used—85% of companies use JVM-based languages for 56% of their application workloads. This usage remained consistent with previous years, demonstrating the lasting and widespread predominance of JVM-based languages across industries.

Go is a fast follower, with 81% of companies using the language for 19% of their application workloads. Similar to the rates associated with JVM-based languages, the adoption and utilization trends of Go and other programming languages remained mostly consistent in 2024.

These stable rates indicate the continued preference for proven and scalable approaches to support the unique and robust needs of cloud-native solutions.

As organizations develop greater technical proficiency with Kubernetes, they can expect increased stabilization, predictability, and standardization, along with opportunities for optimization and innovation. Rapid advancements in AI along with the growing security concerns will drive the evolution and adaptation of Kubernetes usage and monitoring to meet modern business needs, paving the way for new trends.

However, Kubernetes implementation alone does not guarantee increased business value. To maintain Kubernetes health, security, operational efficiency, and cost efficiency—and to drive optimization and innovation—comprehensive Kubernetes observability with AI-driven analysis is essential.