The Devops Chicken or the Agile Infrastructure Egg?
Source – itbusinessedge.com
Does devops lead to agile, or does agile lead to devops? Or perhaps they move in tandem as the enterprise gropes its way through digital transition. And if that’s the case, is optimized, automated infrastructure the cause or the effect of this new IT model?
The answers to these questions could be crucial for the enterprise over the next few years because they speak directly to how technology decisions will be made. For instance, if the right infrastructure is required for devops, then what technologies are needed to deliver the appropriate outcomes? But if devops evolves naturally, then how does the enterprise foster an integrated IT environment rather than simply another collection of disjointed point solutions?
According to a recent survey by BMC Software, the top three priorities for IT investment over the next two years are containers, workload automation/scheduling and devops. But exactly how these tools will work together in what is likely to be a distributed data ecosystem is still a mystery, particularly when organizations are also struggling with budgetary constraints, a skills gap and an extremely tight timeframe in which to make it all happen. Indeed, many respondents to the survey indicate that the lack of a clear path going forward is not driven by technology issues alone but also by conflicting objectives among business leaders and the need to realign internal organizational structures to take advantage of things like rapid and continuous development.
Indeed, say McKinsey & Co.’s Thomas Delaet and Ling Lau, many organizations that have implemented devops in the cloud and on reconfigured internal infrastructure often find that they hit a bottleneck when they reach the point of extending their workflows to legacy environments. With much of the enterprise still stuck on the “plan-build-run” operating model, more agile modes of operation must overcome substantial cultural inertia in order to make a real difference, only part of which is achievable through advanced infrastructure. This puts many organizations in a catch-22 because they can’t disrupt too many processes too quickly, but at the same time cannot afford to move too slowly while more agile competitors eat into market share.
Ideally, the development of a devops culture and its underlying infrastructure should move in tandem, but only in cases where there is a legitimate business reason, says Valentin Bercovici, cofounder of automation and artificial intelligence start-up Peritus.ai. In a recent interview with Silicon Angle, he said this can be a difficult transition to manage because it requires a more formalized approach to devops to replace the skunkworks structure that exists at many organizations. And in the end, it usually increases the power and influence of development teams over non-technical aspects of the business like budgets and market development because they are the ones creating the services and APIs that drive revenue. Just ask today’s business managers how they feel about that.
Many organizations are attempting to square these circles by reserving devops and other advanced architectures for the cloud while maintaining traditional operations on legacy infrastructure – what Gartner terms “bi-modal IT.” But this is a false choice in most cases, says tech consultant David Linthicum, because neither mode provides full support for devops processes. For instance, most public clouds support pipeline management and continuous integration services but are lacking when it comes to continuous test and deployment. As well, the cost of devops in the cloud is often higher than expected, particularly when processes incorporate traditional resources and systems. And for all the talk about automation, the fact is that even the most sophisticated systems cannot provide the full-stack control needed to support an effective devops model in the cloud.
This doesn’t put the enterprise in a very enviable position. It seems that devops can only exist within a completely optimized environment, but achieving such a state will wreak havoc on existing business culture and processes – to the extent that many organizations might not even survive the transition.
And yet, the die has been cast. Digital services are quickly becoming the revenue-generators of the new economy and devops is the only way to provide the necessary flexibility to ensure those services remain relevant and accessible for the consumer.
Hopefully, and sooner rather than later, someone will come up with a general outline for accomplishing this feat. But in the meantime, organizations should plan on pushing devops as deeply into their business models as possible, and then think of ways to overcome the resistance that will hamper deployment at each subsequent step.