DORA (DevOps Research and Assessment) metrics measure how an organization delivers software and recovers from failures. For a platform team, they provide feedback on whether changes to delivery tools and workflows are helping application teams deliver faster and more efficiently. section 9.0 section 1.4.3
The four software-delivery metrics cover speed and stability:
- Deployment frequency measures how often software is successfully released to customers. section 9.1
- Lead time for change measures how long an application-team change takes to reach live customers. section 9.1
- Change failure rate tracks problems introduced by changes to production. section 9.1
- Time to restore service measures how long recovery from a production problem takes. section 9.1
Calculating these metrics is an integration problem: observe the systems involved in delivery, capture relevant data, and aggregate it. An off-the-shelf calculator is useful only if it can accept the systems your teams use. A Kubernetes-based approach can instead standardize incoming information so that sources and metric calculations remain extensible. section 9.1.1 section 9.1.1
A deployment-frequency pipeline
For a deployment-frequency calculation, incoming CloudEvents can be stored in PostgreSQL, transformed into CDEvents, and stored in a separate table. A periodic calculation stage reads unprocessed CDEvents to produce metrics. Retaining the incoming data also allows transformations and calculations to be rerun if calculated metric data is lost. The same stages can accommodate new sources and domain-specific metrics. section 9.2.2
DORA added a fifth, reliability-focused metric in 2022. The four metrics above remain the focus when assessing application-team software delivery rather than broader operational performance. section 9.1
