I had the chance to attend a new conference on the DevOps circuit called DevOps Enterprise earlier in the month. For those of you who did not have a chance to attend, it was a conference co-hosted by one of the greatest Tech Connectors to have ever walked this Earth, Gene Kim, his colleagues at IT Revolution Press and the main sponsor Electric-Cloud. The timing of the conference was spectacular as it happened to fall around the five year anniversary of the DevOps movement.
I first learned about the Conference from a friend and colleague in the Washington, DC area named Jeff Gallimore, a longtime friend of Gene Kim and a collaborator on theDevOps Defense Audit Toolkit. Jeff had a speaking role at the conference with a few auditors and technologists, acting as the moderator on a panel about the Toolkit. Simon Storm, Josh Corman and Byron Miller participated in the panel and have been part of the Toolkit project for quite some time. I won’t go into too much about the Toolkit, other than to suggest reading the blog post and joining the Google Group if you have an interest in participating.
When I first thought about going to conference I was a little weary of what the topics would be like and who the attendees would be. The thought of DevOps concepts and culture inside the enterprise wasn’t something I naturally aligned. As a long-time Velocity Conference attendee, I was concerned that this conference would miss the mark and not teach me something new about the movement or introduce me to new ideas and practices. I personally watched the DevOps movement come out of Velocity, long before the term DevOps was even coined, so any other conference that was going to use DevOps as a culture would have to be top notch. Since Gene was running it, I gave him the benefit of the doubt.
Then I thought about my new role as VP of Engineering at Contrast, specifically about who are our customers. The bulk of our customers are not Silicon Valley startups or technology plays. Rather, they are enterprise players. Naturally, I thought it would be a good idea to see how the “Enterprise” was thinking about the DevOps movement and how they associated it with building and operating software system.
Thoughts on the Conference
I wrote a ton of crazy notes about each of the sessions from Day 1. I guess you can call me the “King of Horses” then. There was one session that stood out in my mind more than anyone and that’s the Target presentation. Ross Clanton (Ops) and Heather Mickman (Dev) presented about how the DevOps culture was brought into the organization at scale. It was how Dev + Ops found a home together at a large Enterprise player with years of legacy code and systems. You can also see the slides here.
Target is/was/will always be a big horse. They are doing the things that make them unicorn like to their competitors and their colleagues in other companies. What are those things you might ask? Well, first they are breeding a culture of transparency through a few core means. The first is making the move to collaborative development via Git using Pull Requests. Everyone uses Git on both the Dev and Ops side. They share repos and access across teams. Second, not only are they big players iMinneapolis’ DevOps Days, but they are running their own internal DevOps Days so that everyone in the organization can participate. Third, they are sharing to the outside world little by little via their Github blog.
The most impressive thing I waked away from the talk was this notion of Flash Builds. As they described, Flash Builds =
(flash mob + scrum + hackathon = awesome) 8hr day with 2x4hr sprints (includes retrospectives, planning, etc…). They promised a blog about the idea. Since it’s on Twitter and the Internet it has to happen ;). From their description, it sounds like Flash Builds are mini-hackday events that accomplish a full development and operations lifecycle. What this team showed as well as anyone is the need to breakdown silos and become one culture.
Devops More than Tooling
One of the pleasant surprises of the conference was the alignment of DevOps and Continuous Integration. I have to admit that I was totally expecting a ton of talks talking about the token DevOps tool sets for automation (Chef, Puppet, Ansible, Salt, etc…). I was expecting a ton of Docker, Vagrant, AWS Services and of course monitoring/reporting (Graphite, StatsD, CollectD, New Relic, etc…). The tooling was called-out in a few presentations, but it wasn’t three days of “Tool Overkill” and presentations about those tools.
I definitely felt like the enterprise players put a great amount of emphasis on build pipelines, feedback loops to development teams and collaborative deployment architectures involving development and operations working together. There were more references to Jenkins-CI than any other tool, followed by Git. In my mind that’s telling me the crowd was focusing on the development delivery pipeline striving for continuous development and feedback.
Unicorns, Horses and Ninjas
One of the main themes of the conference is the notion of horses and unicorns. The companies and players associated with the early DevOps movement were/are often referred to as Unicorns, as their work is considered magical and mystical. This has been echoed in papers, blogs and countless slides at DevOps Days, Velocity, FlowCon, OSCon and other conferences. It was fitting that Gene used this theme as I believe he calls out Project Unicorn in The Phoenix Project, a must read for any and all.
During the conference I found myself writing down some notes on what I thought a Unicorn looked like in the movement, as well as a Horse so I could put a picture to it. I even added a third persona called a DevOps Ninja. I’ll briefly post my comments on the three below. Note, I’m not intending to be snarky, but it reads pretty snarky.
Unicorns are companies that simply get it. They have broken down the culture barriers. There are no silos. Everyone practices Continuous Delivery, They post all of their awesome tools that they built in-house on Github. When they don’t build their own tools, they make time for Pull Requests on the latest and greatest projects on Github. They attend all of the latest conferences as presenters and sometimes as sponsors. Their most favorite part about conferences is the “Hallway Track”. They never sleep. It’s almost like they are vampire unicorns or something. If they aren’t coding on a plane, attending a conference or participating in a hack-event for a declining population of siamese albino wales, then they are guest speakers on podcasts. When do they have time to sleep or better yet…play golf?
What exactly are horses then? I guess it’s fair to say that they come on to the movement well after the movement is gaining steam, but rather has no become a fad. Why? Because they have been heads down on some important projects for the last few years and haven’t had time to pick their heads up. They manage their own source code (SVN, TFS or Perforce). They buy enterprise monitoring tools from IBM, HP and CA. They run their own data centers. They choose Perl and Shell as their scripting languages of choice. They go to conferences, but tirelessly take notes and even pictures of the slides on their smartphones and tablets. They don’t like to get up from their seats at the conference in fear of losing their spot.
I decided to add a 3rd persona in the mix. Let’s call them DevOps Ninjas. Who are these ninjas we speak of? Well, they are the guys from Docker. Ah…just kidding. Well they might be from Docker, but essentially these are the guys and gals who skip the conference all together and have an “Un-Conference” or sponsor secret meetups. They are full-stack engineers who code-automate-deploy-fix-redeploy. They work purely in the cloud. They drift from Starbucks to Starbucks daily.
Should We Aspire to be Unicorns or a DevOps Ninjas?
I think it’s fair to say that there are no such things as unicorns or ninjas. They really don’t exist. I know a lot of folks who work for companies that are considered both and between you, me and this blog post they all have issues. Don’t get me wrong…they love their jobs. They feel empowered to make decisions. They collaborate a ton. They don’t live in bliss. Heck, watch a episode of My Little Pony and you will see that even the unicorn ponies on that show have drama. (Note: I am the father of two daughters ages 8 and 5, so we have My Little Pony on a queue at our house).
I guess it’s fair to say that all companies can be compared to horses. You have ponies, jumpers, thoroughbreds, clydesdales, etc…Companies can be any or all at the same time depending on their attitude and culture, probably the best takeaway from the conference in my opinion. It makes me think of a great post by John Willis back in 2010 called What DevOps Means to Me. In the post, Willis references C.A.M.S. (Culture…Automation…Monitoring…Sharing), which many in the community identify as the four pillars of DevOps. I think the conference presenters nailed the C and S. Nearly every presentation focused on Culture and Sharing. Throw in the word Collaboration for C.S.C and you have a more robust picture of the mood and attitude of the presentations. Collaboration was truly being called-out between the traditional roles of Dev and Ops.
I think the Enterprise represents the line of demarcation between Development and Operations so well. For years you often had the development folks working in cubicles in the bowels of the corporate headquarters. The Operations folks they worked in some NOC or Data Center miles, sometimes many states away. Most of the time the developers didn’t know the operations folks and the operations folks didn’t know the developers. What the conference highlighted and the DevOps movement highlights so well is that the barriers have to be removed. The two cultures have to become an amalgam of one culture. These groups need to be together in order to breed a new ideology of thoughts and perspectives together.
Would I Go Back in 2015?
Absolutely…but I want to go back as a presenter. I felt like there were gaps in the stories I heard. The biggest gap is that Performance and Security continue to be called-out as something important to the Continuous Delivery pipeline, but it’s only lip service in my mind. The only folks who are making strides with Performance and Security remain to be the Operations folks. Incorporating Performance and Security as part of the build pipeline is not happening with the diligence and intensity that it should. These are by far the most important non-functional requirements of any system or application. Yet, they remain to be treated as second class problems by development teams.