The rise of platform engineering doesn’t mean the demise of DevOps, but it could bring an end to the tool sprawl that DevOps unleashed, a KubeCon Europe panel declared.
Platform engineering has become the topic du jour in dev circles. Even the latest State of DevOps report from Puppet said companies who are getting the most out of DevOps are those that have adopted a platform approach, by “designing and building self-service capabilities to minimize cognitive load for developers and to enable fast flow software delivery.”
HashiCorp EMEA field CTO Sarah Polan said that with DevOps, “We wanted to make sure that we were enabling developers to build and run their own infrastructure.”
But, she continued, “Then we quickly became cognizant, both from a business standpoint, but also from the technical and developer standpoint, the developers probably didn’t actually want to run and own their own infrastructure.” Moreover, having each team run their own infrastructure, simply wasn’t scalable.
Reducing cognitive overload and smoothing context switching allows devs to concentrate on development, said Red Hat’s director of market insights for hybrid platforms, Stu Miniman.
This is particularly important in the current skills market, said Polan, where attrition rates are over 20 percent. “If you’re having each of these people string together individual bespoke solutions, and then that person leaves, where does that leave you in terms of business to be able to continue growing and continue innovating?”
But if platform engineering is about removing the drudgery and cognitive overload from human developers, it stands to reason that at least some of this might be better handled by machines, rather than other humans.
AI or machine learning already makes perfect sense when it came to digging into data for trends or incidents. “I absolutely see the role of generative AI when I’m talking about platform engineering,” said GitLab chief product officer David DeSanto.
Going further, he said, “If you’re talking about it from a chat-based interface, you could get into a situation where like, Hey, what is one of my problems going to be, based off the latest usage? Or, hey, I need to apply a patch to this area, what is the best time of the day to do that?”
But Polan also raised the obverse point, asking how we ensure AI can unlearn, when necessary. “So if we give it something that we consider to be best practice, that turns out that’s not best practice? What are the implications of that further on down the road?”
DeSanto offered some reassurance for those who rankle at the idea that their financial and emotional investment in DevOps over the last decade has been nullified by the rise of platform engineering. Platform engineering is about making sure that DevOps and DevSecOps teams are more effective.
“I sometimes struggle with why something has to go and die for something else to exist,” he said.
It’s not like, “one has to be Yoda and the other one Darth Vader.”