AI Speculation Dominates Cloud Native Conference

Talk of AI dominated the official program for the first day of KubeCon this week in Chicago, IL.

Other topics were buried under layer upon layer of breathless enthusiasm for a nebulous term that is rapidly losing all meaning as every technology company scrambles to slap a Now With AI! sticker on whatever they were already doing. The CNCF proudly demonstrated how Kubernetes could be used to build a chatbot with a pre-trained large language model. Much was made of how Kubernetes and containers were in widespread use by companies surfing the wave of hype around this technology that has dominated headlines for almost a solid year.

The drive to talk up one’s AI credentials can be partially explained by media and analyst fixation with the concept. There were precious few questions about anything else from those tasked with interpreting and communicating technical topics for their audience. It is hardly surprising that companies will respond to this pressure to have something to say about AI, however tenuous the link might be.

This author finds the whole thing rather frustrating.

In conversation with actual customers and vendors at the show, AI rarely comes up. When it does, it comes as a formulaic bromide, a shibboleth to signal one’s awareness of the current fashion before moving on to other matters.

The focus of customers is on more practical concerns like how to operate and secure systems built with Kubernetes, which remains a somewhat complex beast. Platform engineering has emerged as the dominant descriptor of what customers are trying to assemble: a shared service that developers can consume much as they once consumed public cloud.

Interestingly, Intuit noted that their implementation of DevOps, where developers are also responsible for production operations, wasn’t working very well at scale. The company is moving back to an approach with dedicated operations people responsible for keeping an infrastructure platform running. A platform that developers use, but don’t have to build or run themselves. The pattern is repeated at other well-known companies such as Discover, Boeing, and Cruise.

Kubernetes has now matured sufficiently that the concern is less with how to adopt it and build the first few clusters than with how to wrangle the dozens, or hundreds, of clusters that now exist. We are firmly in the operations and maintenance era as customers try to settle down and rationalize their approach after years of innovation and sprawl. The shift in wider economic conditions has helped turn people’s minds to efficiency and simplification, which sits in stark contrast to the boisterous pioneering mindset of those keen to boost AI.

Companies such as NetHopper and Rafay are working to make wrangling Kubernetes easier for operators with fleets of Kubernetes to manage. Others, such as PerfectScale, help customers right-size their infrastructure by finding and fixing waste and bottlenecks. Security remains a challenge. Projects such as Falco and TUF are providing open source options that integrate with the CNCF ecosystem, but there is still much work needed to improve the security of the software supply chain.

The CNCF now has 173 projects in its care, at various stages of maturity. The vast majority are concerned with more prosaic practicalities of building and operating important infrastructure. As CNCF executive director Priyanka Sharma reminded us during her keynote, open source projects underpin nearly everything we do in our modern, interconnected society. It’s nice to see that most of them are focused on doing this well, not on chasing the latest fad.