Service Delivery: The Build Phase - The Practice of Cloud System Administration - page 199

leases keep automation fresh and encourage us to update it to reflect small changes
before they turn into major ones.
• A Culture of Continuous Improvement: The ideal process is always evolving
and improving. Initially it might have manual steps. That's to be expected, as pro-
cesses are malleable when initially being invented. Once the end-to-end process is
automated, it can be instrumented and metrics can be collected automatically. With
metrics we can make data-driven improvements. For a process to be continuously
improved, we not only need the right technology but also need a culture that em-
braces change.
• Improved Job Satisfaction: It is exciting and highly motivating to see our
changes rapidly put into production. When the interval between doing work and
receiving the reward is small enough, we associate the two. Our job satisfaction
improves because we get instant gratification from the work we do.
Rather than focusing purely on cycle time, a team should have metrics that balance the ve-
locity of individual aspects of the software delivery platform. We recommend that every
DevOps team collect the following metrics:
1. Bug lead time: Time from initial bug report to production deployment of fixed
code.
2. Code lead time: Time from code commit to production deployment.
3. Patch lead time: Time from vendor patch release to production deployment.
4. Frequency of deployment: How many deployments to production are done each
month.
5. Mean time to restore service: Duration of outages; from initial discovery to re-
turn to service.
6. Change success rate: Ratio of successful production deployments to total produc-
tion deployments.
9.3 Build-Phase Steps
The goal of the build phase is to create installation packages for use by the deployment
phase. It has five steps:
1. Code is developed.
2. Code is committed to the source repository.
3. Code is built.
4. Build results are packaged.