With the rush to cloud enterprise comes increasing use of Kubernetes to get applications up and running on the web. A recent study by big data monitoring firm Pepperdata looked at both the growth of Kubernetes use and how companies are addressing it from cost and revenue fronts.
Pepperdata’s The state of Kubernetes 2023 report found that, on average, organizations deploy between three and 10 Kubernetes clusters. It also revealed that the use of the open-source container orchestration system is expanding to data ingestion, cleansing, and analytics, databases, and artificial intelligence and machine learning.
SEE: Hiring Kit: Cloud Engineer (TechRepublic Premium)
Pepperdata, in its survey of 800 C-level execs and DevOps professionals working in financial services, healthcare, technology and advertising, asked:
- How many K8s clusters organizations run.
- Which workload types do they deploy on K8s containers.
- Challenges encountered by enterprises as they adopt Kubernetes.
- How enterprises measure the ROI of their K8s deployments.
- Where companies stand in their FinOps journey.
- Kubernetes: Deployment beyond microservices
- Looking at Kubernetes with an eye on ROI
- Cost surprises a key challenge for K8s
- Organizations walking toward cloud cost reduction
Kubernetes: Deployment beyond microservices is driving broader use
As Kubernetes reaches maturity and becomes an industry standard for container orchestration, its uses are also broadening beyond its core application as a mothership for microservices. The study found that:
- 30% of executives reported having three to five K8s deployments.
- 38% reported six to 10 clusters.
- Almost 15% said they had between 11 and 25 clusters.
- 4% reported having deployed more than 25 clusters.
In terms of how enterprises are deploying Kubernetes for specific workloads, Pepperdata found:
- 61% of surveyed companies are using Kubernetes to deploy data ingestion, cleansing, and analytics through software like Apache Spark.
- 59% are using Kubernetes for deploying databases or data cache via platforms like PostgreSQL, MongoDB and Redis.
- 58% reported using Kubernetes on web servers like NGINX.
- 54% said they are deploying AI/ML software, such as Python, TensorFlow and PyTorch on Kubernetes.
- 48% said they are using Kubernetes for programming languages like Node.js and Java.
- 42% reported using Kubernetes for logging and monitoring through programs like Elastic and Splunk.
- 35% said they are deploying application servers with Kubernetes.
Microservices are still a good proxy for Kubernetes deployment
Pepperdata’s study suggests that organizations will be adopting Kubernetes in greater numbers, given their plans to deploy microservices like NGINX. Forty-four percent of respondents said they plan to do so this year, while 36% said they have microservices deployed already and only 20% saying they had no plans to do so.
Also, the majority of those polled said Kubernetes provides them a strong foundational architecture for microservices, and that it enables applications to be deployed more rapidly and supports platform consistency across development, testing, staging and production clusters.
Looking at Kubernetes with an eye on ROI
Pepperdata discovered that among those polled, cost to deploy was the leading metric for measuring Kubernetes’ ROI, with findings suggesting that almost 44% of the organizations are looking at ways to implement cloud cost reduction.
After cost, top-line growth (54%), resource usage (49%), followed by deployment frequency (48%), developer productivity (46%), infrastructure utilization (35%) and IT staff productivity savings (25%) were key ROI metrics. Firms reported they expect Kubernetes to increase ROI by lowering administration and operations burden, accelerating deployment times and making resource management more efficient.
Cost surprises are a key challenge for K8s
When Pepperdata surveyed IT leaders about the challenges they faced in adopting Kubernetes:
- 57% said significant or unexpected spending on computation, storage networking infrastructures and cloud-based IaaS.
- 56% cited the learning curve for employees to be able to upskill for operations and security in Kubernetes environments.
- 52% pointed to limited support for stateful apps (such as applications that save client data).
- 50% said lack of visibility into Kubernetes’ spending.
Organizations are walking toward cloud cost reduction
In its FinOps performance study, the FinOps Foundation among other things defines the levels of familiarity with FinOps from crawl to walk to run. In Pepperdata’s study, most respondents self-identified at the walk stage.
The study said that nearly all respondents were familiar with cloud cost optimization, while 32% characterized themselves as “crawling.” The majority (43%) said they are “walking,” meaning they have the ability to implement cloud cost reduction recommendations today. Seventeen percent self-reported as “running,” meaning they are actively reducing costs through autonomous procedures. Six percent said they have not started.
Interestingly, more than 98% of respondents indicated familiarity with FinOps and saw themselves somewhere on the continuum of implementing best practices for cloud cost remediation. In addition, more than 17% of respondents identified themselves in the run stage, with the ability to remediate cloud costs autonomously.