The Puppet State of DevOps Report 2017 shows that high-performing IT teams are deploying more frequently and recovering faster. There is more focus on automation, with loosely coupled architectures and teams facilitating continuous delivery. Transformational leadership and lean product management practices are also key drivers behind high performing teams.
This year’s survey, which is conducted annually by Puppet, reached out to about 3200 people across the globe, including executives, developers, and IT professionals. The number of respondent who stated that they work on DevOps teams has increased from 16% to 27% in the last three years, indicating a rise in DevOps adoption. The biggest number of respondents were from organizations with 100-499, 500-1999 and 10k+ employees, with most of them being from the DevOps, IT Ops/Infrastructure and Development/Engineering categories. The gender ratio, however, remains skewed, with females and others accounting for just 6% and 3% respectively. North America had the largest number of respondents (54%) with Europe and Russia at 27% and Asia at 10%. Technology companies still led the way (34%) like last year, followed by financial services (14%), followed by education, retail, telecom and government agencies in the 6-8% range.
The report distinguishes between high-performing and low-performing teams, and lays out the differences in performance between them. Similar to last year, the following performance metrics were measured:
- Deployment frequency - how often deployments are done to production
- Lead time for changes - how speedily can new changes be pushed to production
- Mean time to recover (MTTR) - the mean time to recover from a failure (outage)
- Change failure rate - how often a change leads to a failure in the deployment pipeline
High performers have improved on all metrics as compared to the previous year. They have 46 times more frequent code deployments and a 96 times faster MTTR. On average, low performers compared to the previous year have also improved on multiple metrics.
There is a significant rise in automation practices, especially among the high performing teams - 28% more in configuration management and 26% more in deployments have been automated.
The ability of an organization to do continuous delivery (CD) was measured by two factors - the ability to deploy on demand to production, and having fast feedback loops for everybody in the team. The factors contributing strongly to achieving these outcomes include comprehensive version control, CI, trunk based deployments, including security in the software delivery process, test and deployment automation. The architectural factor which contributed to CD was loose coupling for both services and teams. Coupling between services was measured by whether people can do testing without requiring an integrated environment and if those services can be deployed independently of others.
The non-technical aspects of achieving high performance DevOps include lean product management. The report defines this as three capabilities - splitting work into small batches and making the workflow visible, gathering, broadcasting and implementing customer feedback, and giving development teams the authority to create or change specs as part of the dev process, without requiring approval.
Leadership is not enough by itself to achieve high DevOps outcomes. It also depends on suitable architecture and good tech practices. The report authors used the Structured Equation Model (SEM) as the predictive model to measure relationships between constructs that are relevant in software delivery.
The arrows in the figure represent statistically significant relationships.
All images courtesy : The 2017 State of DevOps Report.