Torsten Volk — Enterprise Management Associates
With over 15 years of enterprise IT experience, including a two-and-a-half-year stint leading ASG Technologies’ cloud business unit, Torsten returns to EMA to help end users and vendors leverage the opportunities presented by today’s hybrid cloud and software-defined infrastructure environments in combination with advanced machine learning.
Many enterprises see IT operations as a necessary evil that is needed to support our laptops, phones, and tablets, make sure websites do not go down, and own the SLAs for dozens or hundreds of enterprise applications running in a compliant manner. When I was still a dev lead, we would refer to the IT team as “the guys who always say no,” but at the same time, we recognized their skill in compliance and security.
The Guys Who Slow Everything Down
The line of business typically regards the corporate IT department as a roadblock. When asked about why a release is late or of poor quality, developers like to (often rightfully) mention that test environments are too hard to come by and that the developers were bottlenecked by a lack of sandboxes for experimentation and feature isolation. They go on complaining about how hard it is to test with production data due to compliance concerns, and that it is impossible to reliably achieve continuous synchronization between development, test, staging, and production environments.
“Especially when launch time rolls around and our QA team, support team, sales engineers, and beta customers all need their own environments to provide final feedback, we often need to choose between giving them what they need or staying on track with our development,” said a development lead of a major U.S. software vendor. “And in addition, creating these environments takes up a lot of time for our own developers too, as we cannot seem to ever use our scripts without modification and manually support the server guys to get an environment working.” In other words, IT operators today are a long way away from becoming crowd-surfing rockstars.
“The moment we decide to change a single line of code, we trigger the entire compliance nightmare over again, where new environments have to be created, we have to generate a new set of compliance reports, and run the entire test cycle again,” says the VP of Infrastructure Management of a major European Telco. “Even if it is just one line of code, this still takes three weeks and absorbs a small team of engineers and architects on our end and on the side of the developers, too. Nobody is willing to risk any shortcuts, as audit compliance is critical.” Enterprise Management Associates (EMA) research shows that changing a line of code can truly cost 30-40 percent of IT and development resources. This tremendous cost incentivises longer release cycles, instead of staying on the pulse of end-user requirements. “Too often, we develop requirements that are already superseded by ever-changing customer requirements or by the competition coming up with better alternatives,” says a very frustrated head of software development of an e-learning platform. “You want to constantly, quickly, and cheaply be able to validate new ideas that ideally originate from direct communication with our customers.”
Accepting Reality: You May Need To Replace Pets With Cattle
Instead of sending frustrated development teams into their own shadow Kubernetes container journey and driving them sometimes to the outer limits of data locality compliance, IT needs to offer a better alternative to “just moving everything to the public cloud.” But how can IT offer something they were not able to offer in all these years? My answer is pretty simple. Instead of trying to turn pets into cattle, IT has to gradually replace the “old pets” with “new cattle.”
Why Is Cattle So Important?
Cattle is all about immutability of IT infrastructure. Every server environment is in exactly the same state, so new applications can run anywhere. This eliminates the need for IT to assign teams to creating applications and often release-specific environments. Each one of these environments requires extensive application-specific knowledge, and development teams love to ask for “special favors” that quickly turn every VM into a snowflake.
Rockstars Start Fresh and Follow Four Key Rules
The moment the first custom application environment is created is the moment when many of the advantages of a cloud-native and immutable approach to DevOps go out the window, and the old way of doing things returns. Therefore, it is critical to “start new” with a greenfield hyperscale environment that works across data centers and clouds and scales in a linear manner. This environment needs to scale automatically and management needs to be conducted on the entire infrastructure, instead of on individual devices. But will this not create more silos?
Rule #1: Rockstars Integrate, Integrate, Integrate
When you create new, immutable hyperscale infrastructure right next to your old pets, integration and a clear transition plan are critical. This means the new environment must include excellent connectivity to the rest of the data center to enable new cloud-native applications and microservices to tap into legacy APIs. At the same time, the new infrastructure must seamlessly integrate with all popular DevOps tools.
Rule #2: Rockstars Automate
By definition, continuous delivery of software requires full automation of the entire DevOps process, from build, to test, to deployment, to release, to feedback. Only when continuous release is a “state of mind” instead of a “state of emergency” can the organization focus on experimenting with innovative solutions to business problems.
Rule #3: Rockstars Make No Exceptions
The moment we implement the first snowflake server is the moment that leads us down the path to the gates of Mordor, while sending the four headless horsemen right on our tail. Immutability still needs to enable the use of standard databases and environmental variables, but what it cannot mean is the requirement for special software on “just that one server.”
Rule #4: Rockstars Help Unlock The “Other 50 Percent” Of Application Workloads
Finally, true rockstars need to be able to offer hybrid cloud. This means that simply securing public cloud offerings is insufficient, since today approximately 50 percent of applications cannot be moved to the public cloud for compliance and security reasons. Connecting public and private cloud with one central governance and automation overlay is what true rockstars do. That, and playing Paradise City backwards.