Last DevOps Cafe Podcast ( Episode 47 ) previewed the QCon London 2014 DevOps track . The track hosts and most of the speakers talked with Damon Edwards and John Willis , the DevOps Cafe hosts. Manuel Pais and Shane Hastie , the track hosts, explained the rationale behind the track's session selection, the speakers introduced their talks and there was still time to discuss other topics, such as the importance of the scientific method and how agile's definition of "done" must be adapted in a DevOps world.
According to Manuel, QCon is more of a developer-oriented conference so they wanted the track to help build bridges to the development-oriented people, explaining what DevOps means and how it can make a difference for them. The sessions will take different points of view on DevOps: from the culture side of things to the business perspective, while some case studies add concreteness to the whole topic.
Daniel Schauenberg 's "Development, Deployment & Collaboration at Etsy" talk is about how far one can go when there is an intrinsic DevOps culture in place, coupled with Continuous Delivery practices. The talk will show how Etsy's stack is set up: against what one might expect, they are not on the cloud and have a monolithic application. Daniel will highlight how Etsy approaches development, deployment, incident response and collaboration over a common goal.
Damon Edwards ' "Dev 'Programming' Ops For DevOps Success" talk will focus on the culture side of things. Although some organizations can build a DevOps culture from the ground up, there are many that are not so fortunate. As Damon put it, "Enterprise is the legacy of success". Many enterprises have a long list of legacy applications and have built silos over the years. Damon's talk will take a two-pronged approach to help in this context. First, it will focus on the culture side, with tips and tricks on how to bring about change and bridge the gaps between Dev and Ops. Second, it will focus on the need to build a model of how the operations team should work. Operations-as-a-Service is a term that is used to describe an approach where operations activities are not managed through a ticket-based, request-response queue, but act more like a service to be consumed, in a self-service, programmatic way.
Dave Farley 's "The process, Technology & Practice of Continuous Delivery" talk will highlight some of the traits that Continuous Delivery and DevOps have in common: the tooling; the need for collaboration; the short feedback loops. Dave Farley believes that agile works because it establishes short feedback loops, which promotes fast and continuous learning, something that DevOps also advocates. At DevOps Cafe, Dave drew a parallel between DevOps, Continuous Delivery and what he calls "the greatest invention of mankind", the scientific method. The talk will use a real world project example to explain how they used that reasoning to reach Continuous Delivery, why it works and how it works.
Robert Benefield 's "Instrumenting Your Business For Success with DevOps" will be about how the operations team can help the business side see the value in DevOps. Robert believes that good instrumentation is key for the business to understand what's happening in the real world, and how that situational awareness can be a differentiator. Robert also mentioned that the way things are presented to the business is important: "100 deployments a day" does not mean much. What the business wants to know is how fast they can respond to a relevant business event.
Ola Ellnestam could not attend the podcast. Shane explained that his talk, "DevOps at a small International Bank - Contiguous Improvement over Continuous Delivery" , will be a case study on how DevOps can be applied even on a context of tight, complex regulations and where compliance is critical: a bank. Ola is working on a Sweden bank where there's a focus on working on things that add value and it's safe to experiment. In the process they coined a term "Contiguous Improvement", taken to mean, "only improve what needs to be improves". The talk will have lots of stories to highlight the underlying principles and values behind the business and technical choices they made.
The last third of the podcast touched on more general topics, such as feedback loops and agile's definition of done.
Everyone agreed that short feedback loops and the verifiability that they bring is key to a successful DevOps practice, again linking with the scientific method process: postulate an hypothesis; make an experiment; prove or disprove the hypothesis. Enable this cycle to be as fast as possible and do it continuously.
There was also a discussion of the agile definition of "done". There is a concern that this definition can be distorted, letting a development team think that the work is finished once the development is finished and approved by the business, forgetting the operations side. Especially in a services world, nothing is ever completely done. Damon mentioned the term "WaterScrumFall", which describes scenarios where the development team is agile, but up and down the value stream, the business and operations keep working the same way they always did. DevOps was born out of a need to break this line of thinking and bring together all the participants in the value stream.