DevOps is facing a shakeout in the near future, according to Forrester Research. It predicts that businesses will pivot away from release cadence as a measure of how DevOps teams are performing, and instead place more importance on metrics such as improved sales and customer satisfaction.
For developers who always thought that pursuing speed of release was a blind alley, this might seem like vindication. But Forrester doesn’t think that DevOps is going away, rather that it is nearing an inflection point where adoption will change, firms will implement different metrics, and testing will move to include “previously radical methods”.
The report, Predictions 2020: DevOps, claims that multiple Forrester surveys have found release velocity among firms with a DevOps strategy to have stagnated, and that this holds true across industry sectors.
Part of the problem is a mismatch between the rose-tinted perceptions of managers who claim their organisation “actively embraces DevOps” and the reality reported by developers at the coalface. Efforts to “bolt on” Agile and DevOps have failed, and frustration with current approaches will reach a boiling point in 2020, Forrester believes, meaning that a wholesale change of direction will be called for.
The report states that businesses are coming to realise that velocity is a measure of output and not outcome. Speed of release may be a good indicator of the level of DevOps maturity in an organisation, but doesn’t indicate whether investment in speed is worth the effort. In other words, businesses are starting to ask if faster is really better.
Forrester predicts that 2020 will see more widespread use of tools that measure performance and value, such as application performance management tools that can capture real-time performance metrics, and a focus on capturing metrics such as customer satisfaction.
Another prediction is that at least one Fortune 500 company will adopt chaos engineering as a risk control strategy. Readers may think that chaos engineering is what their own company already does, but it is in fact a way of testing the resilience of a system by randomly causing failures.
This previously radical method was pioneered by Netflix to test out its distributed delivery system. It was developed as a response to the increasing scale and complexity of such microservices environments, which are difficult to fully test using conventional methods.
Forrester also believes that DevOps toolchains will increasingly have to feature embedded learning facilities, to enable developers to constantly learn new skills. This is because traditional learning tools such as books, online training, videos and classrooms are separated from the tools developers actually use. Embedding learning facilities into the tools should expedite the process.