A global survey of 501 IT professionals published today by the Data on Kubernetes Community (DoKC) consortium finds organizations are now running more complex analytics (67%) and artificial intelligence (AI) and machine learning algorithm (50%) workloads on Kubernetes clusters.
The survey, conducted by the market research firm ClearPath Strategies on behalf of DoKC, finds the most widely deployed stateful platform on Kubernetes clusters is, not surprisingly, databases (76%).
In addition, the survey finds that more than half of respondents (51%) are running more than half their data workloads on Kubernetes clusters.
DoKC director Melissa Logan also notes that one-third (33%) of respondents saw their productivity increase two-fold or more as the management of compute and storage becomes more unified. Respondents cite streamlining their IT operations by ensuring consistency (43%) and simplifying management (41%) as top reasons for deploying stateful applications on Kubernetes clusters.
Benefits of running stateful applications cited by respondents include ease of scalability (37%), the ability to standardize the way all workloads are managed (36%) and consistency across developers and production environments (33%).
However, challenges remain. The lack of integration with existing tools (35%) followed by lack of qualified talent (34%) and the time and effort required to manage (31%) are all cited as issues by survey respondents. Two-thirds of survey respondents (66%) are currently using 20 or more Kubernetes operators to manage and invoke Kubernetes APIs at a higher level of abstraction.
Well over a third (36%) of respondents say they expect operators to handle all Day 2 operations such as application and storage life cycle, backup and recovery failure, while 40% said they also expect advanced capabilities such as the ability to provide metrics, alerts, log processing and workload analysis.
REspondents say they evaluate security, ease of use and maintainability when looking for a Kubernetes Operator. Nevertheless, more than half (53%) say automating application provisioning and configuration remains a challenge. In the longer term, the issue may soon be that organizations simply have too many Operators to deploy, manage and maintain. In the meantime, 70% of respondents say they expect to either reskill existing members of their IT staff or hire additional specialists to manage data on Kubernetes clusters.
Logan says the survey makes it apparent that the debate over whether or not to deploy stateful applications on Kubernetes clusters is all but over. In fact, the survey finds a full 83% of respondents attributed more than 10% of their revenue to stateful workloads running on Kubernetes clusters. That percentage is only going to rise as more AI applications are deployed on Kubernetes clusters, she adds. In addition, it’s only a matter of time before more stateful applications are deployed on Kubernetes clusters at the network edge to process and analyze data at the point where it is created and consumed, notes Logan.
It’s not clear how much the rise of stateful applications is being driven by greenfield applications being deployed by IT teams that don’t have access to legacy storage resources. However, it’s clear that the next era of the hyperconvergence of compute and storage is being driven by Kubernetes clusters. The primary challenge will be training IT teams that have, in many cases, already deployed hyperconverged infrastructure (HCI) to run virtual machines to now also manage Kubernetes clusters.