Microservices are evolving from fad to trend, according to "The State of Microservices Maturity" survey, conducted by O'Reilly and Neal Ford in July 2018 and published in December. That's a conclusion aligned with InfoQ's own Architecture and Design Trends Report, released recently. The report showed an overall positive attitude towards microservices among the practitioners surveyed for the report. Among the most significant findings in the report is that DevOps and microservices feed off each other, so that the success of one contributes heavily to the success of the other.
Microservices are a success
With 866 respondents, the survey looked at how practitioners implement microservices. While it's a "survey designed for companies engaged in microservices", the first of the key findings is how those companies that have adopted microservices continue to do so with some enthusiasm. Over 50% of respondents say that more than 50% of new development in their organisation utilises microservices. That says a lot about the success of microservices in those companies. That is further backed up by the fact that 86% of respondents described their microservices venture as at least partially successful.
DevOps practices in harmony with microservices
The report highlights the continued push towards DevOps practices amongst microservices practitioners. The majority of respondents indicated adoption of DevOps practices, such as continuous deployment and deployment pipelines including automated testing. According to Neal Ford, this is necessary for the survival of microservices. "Mature microservices architecture requires at least some maturity in DevOps practices." He suggests and that the "....synergy between architecture and DevOps is one of the super powers of the microservices architectural style because it delegates responsibility more intelligently."
That last quote fits nicely with the findings within last year's excellent book Accelerate by Nicole Forsgren et al. The book reports on the state of DevOps practices amongst high-performing teams. In one chapter, it finds that high-performing teams (those that were using DevOps practices) were more likely to be building custom software using a microservices architecture. Interestingly, they also find that a loosely coupled architecture, in the microservices mould, was actually the biggest contributor to continuous delivery.
Kubernetes still an outsider
Kubernetes has had virtual ubiquity in blogs and conferences over the last couple of years. Companies such as Google, Microsoft and Pivotal have continued to put significant investment towards it. And yet this report finds limited adoption of Kubernetes, with over 60% of respondents saying they have yet to use it despite the report finding the vast majority using containers in their development. According to Neal Ford, the survey was conducted while "Kubernetes was in its infancy" and it takes large organisations a while to "turn the large ship of enterprise towards something as disruptive as...other much hyped technology". However, he says the survey is not able to predict the future market for Kubernetes (or its competitors), but acknowledges that "interest in Kubernetes remains" and that the report may not reflect the "highly dynamic nature of the current reality", almost half a year later.
Establishing appropriately decoupled bounded contexts a work in progress
Sam Newman, in his venerable book Building Microservices, dedicates a section on how to use bounded contexts, a Domain-driven design concept, to identify appropriate microservice boundaries. The O'Reilly report, however, found that the level of success in using DDD to identify services was modest. Almost 50% of users considered it a work in progress, with another third of them considering it moderately successful. The report was not looking for adoption levels of DDD but rather "more about tracing the association between the two (microservices was clearly inspired by DDD)", with Neal Ford suggesting in his experience "organizations struggle with finding the correct granularity for domains/services, and the answer seems to bear that experience out".
The O'Reilly source survey occurred over a couple of months, ending in August 2018. This report is the first time it has appeared and was inspired by similar reports in the past, such as the DevOps report for the Velocity conference.