Survey Surfaces Recent Spike in Kubernetes Costs
A survey of 178 IT professionals that have deployed Kubernetes finds just over two-thirds of respondents (68%) have seen an uptick in Kubernetes costs, with an equal number either relying on monthly estimates (44%) or doing nothing to control those costs (24%).
More than half of respondents report they saw a 20% increase in Kubernetes-related spending in the last year. Only 12% report they have lowered Kubernetes expenses during the same time period, while 20% report they have been able to keep costs relatively constant. Only 10%, however, said their Kubernetes costs currently exceed $1,000,000 per month.
Conducted jointly by the Cloud Native Computing Foundation (CNCF) and the FinOps Foundation, both arms of the Linux Foundation, the survey finds three-quarters of respondents are running Kubernetes in a production environment, while 16% are using it in development, test or proof-of-concept (PoC) environments.
The survey finds almost half of respondents (44%) believe predicting cloud bills is very important. However, only 38% report that they can predict monthly bills within a 10% margin of error, while 21% do not or cannot predict their bills. Most Kubernetes-related spending is attributable to computing and memory resources (90%), the survey finds.
Only 13% of respondents say they track accurate showbacks, while 14% said they have a chargeback program in place. Among those monitoring costs, the most widely employed cloud cost tools are from Amazon Web Services (15%), Google (11%) and Microsoft (8%). The open source cost monitoring tool Kubecost is the most widely deployed across multiple platforms (13%), with 11% reporting they rely on homegrown tools.
J.R. Storment, executive director for the FinOps Foundation, says the primary focus is not to reduce spending but rather make sure resources are being used efficiently to enable engineers to build and deploy applications faster. As part of that effort, however, finance leaders are requiring organizations to better predict costs, especially in the wake of the economic downturn brought on by the COVID-19 pandemic, notes Storment.
As the overall economy slowly recovers, organizations are looking to maintain a level of financial rigor in terms of managing application development costs that often wasn’t present prior to the pandemic, adds Storment.
Ultimately, the goal should not be to discourage developers and engineers from innovating by limiting how much resources are consumed, notes Storment. Rather, organizations need to provide DevOps teams with more visibility and transparency to encourage optimal consumption of IT infrastructure resources, says Storment.
The challenge now is putting the guardrails in place that ensure there are no unpleasant surprises any time that Kubernetes bill comes due, adds Storment.
Kubernetes costs, of course, are a function of the rate at which cloud-native applications are being deployed on the platform. However, there are always going to be IT reactionary forces that resist migrating to any new platform. If the cost of managing Kubernetes clusters rises to a level that is perceived to be higher than legacy platforms, then resistance will stiffen regardless of how much more agile an organization becomes. The issue is finding a way to enable that agility in a way that isn’t perceived to be blowing the IT budget.
