Good News: DevOps is not One-Size-Fits-All

Want to try something fun? Ask different groups around your organization what “DevOps” means. It’s fun because the odds are good you’re going to get very different answers, many of which will be vague, if not outright contradictory. This is perhaps unsurprising when even industry-recognized DevOps luminaries tout varying definitions, approaches, tools and methods.

Here’s what’s not so fun: trying to figure out how DevOps might look at your organization, and what resources and amount of time it would take to do it right. The one thing upon which everyone seems to agree is that “DevOps is a good thing” and something organizations ignore at their peril, in light of the significant competitive edge it provides.

Here’s the good news: DevOps is not one-size-fits-all. In fact, there are many different facets in defining DevOps relative to your purpose(s). This article will identify some of the axes along which the answers to those questions fall, and sketch how they contribute to the overall picture. By the end, you should be better equipped to navigate your organization through the resulting, thoroughly non-Cartesian witch-space that is DevOps today.

Axes and Allies

My use of “axes” in the previous paragraph was neither random nor merely picturesque. It was chosen because it’s a bit of a chameleon, the only English word that can be the plural of three separate noun forms. In short, you don’t know what the word means apart from context. The same is true of DevOps.

Linguistic pedantry aside, DevOps grew out of the organizational friction between developers and operations, hence the name “DevOps.” But the DevOps movement pearl unquestionably has been layered and shaped by the oyster of today’s development trends: the increased (if not nearly exclusive) focus on web and mobile applications. This is simply because it’s an easy fit, those products often being much simpler to build and deploy than legacy applications.

This gives us the first axis: the industry (or industries) in which you compete. If you’re building the next great social networking site—please don’t, I beg you!—you’re going to have a tough row to hoe. And you’re going to have to do at at the speed of Web whatever-dot-oh. Because continuous deployment isn’t an option for you; it’s your personal Obi DevOps Kenobi (your only hope). Yet, at the other end of the spectrum, even the most legacy-heavy manufacturing concern can benefit from DevOps principles, though the requisite pace of innovation should be rather lower.

Industry also can determine the degree to which you accumulate necessary but hard-to-satisfy stakeholders. In the medical industry, for example, keeping your compliance department happy is both essential and more than a full-time job in itself. The DevOps drive to deploy quickly will run straight into the sort of bureaucratic inertia one expects in a heavily regulated industry. You’ll likely need to temper your DevOps expectations and shape your processes to deliver and archive additional content if your industry is more legacy than modern.

A second axis around which DevOps revolves is the type and level of maturity of products you deliver. The more you lean toward legacy products, languages and processes, the more you’ll likely need to scope and direct your DevOps efforts differently, depending on the answers to a variety of questions. For example, it’s rather a different thing to build, test and deploy a monolithic C++ application than an app intended for the web or mobile devices.

Ask yourself: How complicated are the true essentials of your build process? Which parts can you tighten? Can you isolate and prebuild components and leverage an artifact management system? What build artifacts must be retained to satisfy support, compliance and other stakeholders relevant for your industry? You’re going to need tools that can handle the load in every sense.

Similarly, how much testing can you realistically automate for the best payoff? And what sorts of testing will give you the strongest guarantee of success for the effort? Unit tests are a good start, but have you automated your all-up functional tests? What about UI testing? Stress testing? Again, the type of industries you serve and the type of products you create will generate different test demands and degrees of time and effort to automate your test suites.

A third axis along which to evaluate DevOps is your organizational culture. I’ve heard “legacy” defined as code lacking unit tests, but this seems incomplete. I’d suggest that, in fact, there’s a legacy state of mind as well, by which I refer to the sort of inertia that accretes in products and organizations over time. You’ll find it expressed in “the way we do things.” If that phrase is a discussion-ender at your organization, then you’re going to have a much more difficult time embracing DevOps.

What allies can you bring to the table to push for necessary change? Attending a DevOps conference or event can connect you with experts with experience in your industry who might be able to recommend materials. The more your organizational attitude “leans toward legacy,” the more you’re going to need a thick bench of allies to make DevOps a reality.

Leaving a Good Legacy

If it seems as though I’ve been hammering on the notion of legacy in all its forms, I have. But it’s important to recognize that legacy isn’t always a bad thing; in fact, legacy applications are essential to a non-trivial fraction of our world. Not every need can be met with a web or mobile app. It can be very difficult to tame the legacy beast in all its forms, but it’s nearly always possible to at least trap it in a well-defined cage, then shrink that cage over time.

DevOps principles can raise your game no matter how legacy your industry, products and personnel are. But it’s important to understand where you fall on all the relevant axes to determine the right scope, methods and tools to deliver the most value. DevOps isn’t one-size-fits-all, but it can be tailored to fit in at your organization.