Serverless, Multicloud Popular with K8s Users, Survey Says

Serverless technology, the cutting-edge tech whose moment always seems to be on the cusp of arriving, has finally arrived — at least for Kubernetes users, according to a new survey.
Eighty-eight percent of participants in the new study of IT professionals said they are either currently using serverless computing or are planning to.
Hybrid and multicloud environments are also being rapidly embraced, according to the study. Sixty-five percent of survey respondents said they are actively moving to that architecture by the end of the year. Another 20% expect to be there within nine months or more.
Not that the transition is easy: One in three survey participants said navigating hybrid and multicloud requirements was their primary concern regarding running workloads on Kubernetes, sometimes called K8s.
The findings surprised Jim Walker, principal product evangelist at Cockroach Labs, which co-sponsored the study with Red Hat OpenShift. The companies launched the survey to discover how organizations that have adopted Kubernetes are using it — and what the obstacles might be to wider adoption.
“Multicloud and hybrid cloud is something that is very difficult, especially with Kubernetes,” Walker told The New Stack. “Typically we think of Kubernetes in one single environment, one single control plane, because of the complexities of networking, storage, security.
“In fact, we’ve been trying to figure out the federation of Kubernetes clusters for quite some time as a community within the Kubernetes community. And it’s been difficult.”
He added that the result showing the widespread adoption of serverless technology is particularly worth noting: “I was really surprised, because I think that’s still kind of cutting-, if not bleeding-edge technology.”
The survey of 202 IT professionals from U.S. enterprises and smaller organizations, representing a variety of industries, was conducted by Gatepoint Research. The overwhelming majority of organizations represented are Fortune 1000 companies.
Challenges: Transactional Workloads
Seven years after the initial release of Kubernetes, lots of questions remain unanswered about how it’s used. Which workloads are running on it? How many? And who’s running them?
The survey sheds light on many of these mysteries, including the challenges users are facing.
K8s was designed to handle stateless applications; stateful or transactional applications, data-rich workloads that save client data from its activities to use in future ones.
Transactional, or stateful, workloads are the most concern most commonly cited by survey respondents, named by 46% of them, in effectively architecting for and deploying with Kubernetes.
“People really want to run transactional workloads on Kubernetes,” he said. “They’re running the database, and it’s next to Kubernetes, it’s just not on it. It’s not in the same environment.”
— Jim Walker, principal product evangelist, Cockroach Labs
The report states that this challenge “was named as a concern by the highest number of participants. It was also chosen as a single concern more than any other category (by a significant margin).” Ten percent of respondents named it as their only challenge.
Walker— acknowledging that, as an employee of a database company, he does have a vested interest in this topic — said he was still surprised by this result.
“People really want to run transactional workloads on Kubernetes,” he said. “They’re running the database, and it’s next to Kubernetes, it’s just not on it. It’s not in the same environment.”
The study found that 59% of participating organizations run both stateful and stateless workloads on Kubernetes, while 28% run or plan to run stateful applications on K8s. Only 15% run or plan to run only stateless workloads on Kubernetes.
Walker pointed to the continuing challenge for teams that want or need to run transactional apps on distributed architecture. “The ephemeral nature of pods, actually, is counterintuitive to state,” he said. “With some advances in Kubernetes, over the last couple of years, it’s become a lot easier. But I think the databases of old were not built for this new infrastructure.”
Among the other most frequently cited challenges:
- Complex migration of legacy workloads was named by nearly 45% of respondents. It ran a close second to transactional workloads, but few survey participants chose this as their sole challenge.
- Training teams and building experts were cited by 41% of survey participants. In addition,
- Hybrid and multicloud requirements were named by 33% of participants.
- Budget was cited by 23%.
- Current bandwidth and hiring were named as a challenge for 18%.
Who’s Watching the Clusters?
Organizations running Kubernetes are using it to either run a handful of services and applications or many, with little middle ground:
- 38% of organizations surveyed are running (or plan to run) five or fewer services/applications on Kubernetes.
- Meanwhile, 42% are running or planning to run 15 or more services and applications.
These results, said Walker and Michelle Gienow, senior technical content manager at Cockroach Labs and author of the report, point to the need for further study, on whether economies of scale play a role in why organizations either stay small or go big, with not much in between, when it comes to using Kubernetes.
Only 5% use fully managed cloud services; the rest rely on in-house expertise; of those organizations that rely on managed cloud services, 75% reported running five or fewer workloads on Kubernetes.
The study also revealed who organizations are relying on to run their K8s workloads.
- Among 54% of participating organizations, a DevOps or site reliability engineering (SRE) team maintains Kubernetes.
- But 41% reported that the responsibility is shared across multiple teams within their organization.
- Of those that use a dedicated DevOps/SRE, 19% say that team has at least 10 members, while 32% have a DevOps/SRE team with two to 10 members.
“What we don’t know is the relationship between these kinds of teams and the kinds of workloads and that’s what we’re going to dig into next,” Gienow said.
She added, “It would make sense that teams, companies that are doing cross-functional teams are typically in a CI/CD workflow, or splitting a lot of workloads among a lot of teams because that’s the whole beauty of my microservices, right? But we don’t know that for sure.”
The results, Walker said, indicate a need for more sharing of information by K8s users on how they’re using the container orchestrator to conduct their day-to-day business.
“The conversation has always been about these technologies,” he said. “It’s got to be more about the real-world use cases and what’s going on out there. That’s the shift. And that’s where we need to be now as a community.”
To learn more about how K8s works in various environments, download for free the latest edition of The New Stack ebook “The State of the Kubernetes Ecosystem.”