My talk at Austin was about looking back over eight years of devops and asking: if I were to hand some information to the next cultural movement, what mistakes did we make and how can we get better? The biggest problem is the technical gravity that always pulls us back from culture toward tools. Every time a new technology hype arrives – containers was the shiny object this time – we somehow lose the cultural and human aspects. People get focused because technology is tangible, something you can try and do. But having a set of weights sitting there does not make you fit.
At DevOpsDays Beijing weeks earlier, they positioned it as three pillars: devops, SRE, and now AI ops. The idea that collaboration assistance does not just come from humans but also from machines – anomaly detection, Clippy-like helpers. Not chatbots, but assistive technology. That human-machine collaboration angle was fascinating.
On the practical side, we had been working with serverless for about a year and a half. We used Lambda to handle burst traffic from live TV events – containers handled the stable baseline load, and overage went through Lambda because the response times were still slightly erratic but the scaling was instantaneous versus the one-to-two minute spin-up of new machines. Real production problems surfaced: deploying a new Lambda version under peak load meant the old version might still be running from an optimized in-memory state. You had to manually flush it.
The principal value of the word devops has been as a label – a way for people to find practices and connect with others doing similar work. People were successful before devops existed. I have no stock options in the term. But the industry is at the beginning of an upward curve where many large organizations are realizing there is a new way to do IT. It will take time for a new cultural term to emerge; in the meantime, we will have many technical floods washing over us.
Watch on YouTube – available on the jedi4ever channel
This summary was generated using AI based on the auto-generated transcript.