William and I were at LinuxCon to give a presentation entitled “Fail often, fail fast.” It tied into central DevOps concepts such as experimentation, constant iteration, and having a culture that supports these types of activities.
Per the abstract, software projects were historically managed on a bet the farm model. They succeeded or they failed. And when they failed (as big software projects often did), the consequences were typically dire for, not only organizations as a whole, but for many of the individuals involved. Today, by contrast, many software and the development projects have evolved toward a much more incremental, iterative, and experimental process that takes cues from the open source model which excuses (and even rewards) certain types of failure.
This presentation talked about how failure can be turned into a positive. This includes the organizational dynamics associated with tolerating uncertain outcomes, the need to define acceptable failure parameters, and the technical means by which experimentation can be automated in ways that amplify the positive while minimizing the effect of negative outcomes.
William and I also ended up backfilling a slot opened up by a no-show presenter. We gave a variant of my “Containers: Don’t Skeu Them Up (Use Microservices Instead
)” talk which gets into how containers can be most effectively deployed together with new technologies and approaches – including the resource management of large clusters with diverse workloads – rather than mimicking legacy sever virtualization workflows and architectures. And how new open containerized platforms, such as OpenShift, provide integration and a place to effectively run new types of applications and application architectures (such as microservices-based ones).