Dev Guys Want Everything Right Now. What’s an Ops Guy to Do?
When software developers demand dev and test infrastructure ASAP and when they ask to release new features once a week or even more often, the business should count itself lucky,since most of today’s differentiation originates from software capabilities that are ahead of the competition. If you can release faster than the others, chances are that you can respond quicker to user demands and offer a better user experience. Years ago, many of us smiled when certain software vendors made bold statements such as “today, every company is a software company.” These statements seemed self-serving and targeted toward increasing software sales.
Maintaining the Status Quo is not a Viable Option
Today, we are no longer smiling, but sweating over solutions that provide developers with the environments and release automation they require to continuously deliver new business features. As an industry analyst, I can tell you my observations of what I see happening if corporate IT cannot or does not want to move at the speed of business developers.
- Shadow Kubernetes: EMA research has shown that 72 percent of enterprises are faced with rogue container environments that are not under the governance of IT. Developers generally prefer Amazon EC2 Container Service, Google Container Engine, and Azure Container Service and will simply shift their applications there.
- Leakage of microservices: In today’s world of microservices (microservices are single functions that serve one specific purpose and communicate via standardized REST APIs), the seepage of applications and their data to instantly available ungoverned public cloud environments can happen quickly. Once it starts, is difficult to under control.
- Developers take over: Development teams that are unable to make their release deadlines are demanding the “keys to the IT castle.” They want to prescribe an automation strategy or even take over the release process. This option is preferable to the above scenarios of shadow containers and microservices leakage, since it prevents the compliance risk that comes with them.
- Drastic outsourcing: This scenario is based on the CEO and CFO putting on pressure to “no longer operate data centers if we cannot do it efficiently anyway.” Vendors such as Amazon, Google, and Microsoft will gladly send their their “Snowmobile” to suck all of the data and applications into their clouds. While this does not have to always be negative, EMA research shows that public cloud cost has been high up on the list of IT pain points in 2017, and it still is today.
Ops Now Needs to Compete with AWS, Azure, and Google
To prevent the snowmobile from taking everything away, corporate IT must offer a consumption model that is as simple as the one offered by hyperclouds. Therefore, IT needs to offer a set of management APIs that integrate right into the DevOps automation pipeline. As long as data center resources are as easily available as their equivalents in the public cloud, developers will no longer have a need to go around IT to get what they need.
In addition to deployment APIs, companies also need to ensure operational continuity, performance, scalability, and load balancing. Finally, a modern data center environment needs to offer the ability to host and manage Kubernetes containers and ideally Functions as a Service.
But How Do We Get There?
Automation silos, unparameterized scripts, and a general lack of automation are all at the heart of the problem of why companies cannot offer flexible self-service in data centers today. Servers still receive their individual configurations based on workload and also based on the preferences of the specific administrator. Network and storage resources are typically tied to the underlying infrastructure instead of being attached to applications. Often, businesses are using different tools to monitor different types of infrastructure, middleware, databases, and applications. In short, there are too many manual steps for the current approach to scale. In other words, businesses need to transition from managing data center resources as infrastructure-centric pets toward managing them as application-centric cattle.