As agile methodology was popularized at the start of last decade, engineering teams gained the opportunity to begin producing software in shorter cycles, while releasing it reliably at more frequent intervals.
This concept garnered greater attention in 2009, when Tim Fitz published his thoughts on continuous deployment. By the time Jez Humble authored his seminal book, “Continuous Delivery: Reliable Software Releases through Build, Test, and Deployment Automation,” in 2010, the approach started to get significant mindshare. In the book, Humble advanced the three building blocks of continuous delivery: configuration management, continuous integration and continuous testing.
Thereafter, Agile Operations—the practice of using end-to-end monitoring, fueled by analytics, to increase operational agility—appeared. This complemented continuous delivery, with the goal of increasing the quality and velocity of software delivery while minimizing overhead.
Since then, and particularly over the last six years as DevOps adoption has exploded, there’s been significant discussion (and confusion) over the opportunity to embrace “continuous everything” across the software development life cycle (SDLC) to deliver functionality continuously and optimize end-user experience.
Today, having spoken to hundreds of organizations employing agile and DevOps, it’s possible to expand on these concepts to better define the key components of continuous delivery.
In this blog, I’ll detail the precepts of “CD” that CA’s customers recognize as critical in reducing the time, risk and expense of software delivery by frequently releasing smaller batches of code with extensive visibility and traceability.
These building blocks of continuous delivery are:
- Continuous development & integration,
- Continuous testing. and
- Continuous release.
Continuous Development and Integration
When it comes to delivering quality apps faster, every second counts.
That’s why it’s important to automate the planning, development, unit testing and integration of code with different inter-dependent components, all while packaging it for quality assurance. We categorize these activities under continuous development and integration, as shown in Fig 1.
This all starts with sprint planning and modeling of requirements in agile management software. Unit tests can be created automatically, based on requirements and use case stories. These unit tests then can be executed automatically based on pre-defined criterion, using automatically generated synthetic data.
Fig 1: Continuous Development and Integration
Next, virtual service interfaces can be generated automatically by simulating unavailable systems across the SDLC, allowing developers, testers, integration and performance monitoring teams to work in parallel—driving faster delivery, along with higher quality and reliability.
The integrated code can be built and tested automatically against validation criterion, and upon passing, made available for continuous testing. This later phase, after code development and before QA, is also commonly referred to as continuous integration.
Continuous testing across app delivery not only enables enterprises to shift testing left, engaging it earlier in the SDLC, but it also helps them shift right, leveraging the same processes throughout staging and production.
After the unit tests are passed and integrated builds are made available for QA, functional and performance tests can be completed—before the build is made available for staging and production. Various activities during continuous testing are shown in Fig 2.
Before the functional and performance tests are performed, the build promotion criterion and configuration also need to be validated. Testers require provisioning of QA environments to deploy the build, along with virtual services and test data. Using test data management to create a subset of production data and mask sensitive information is another continuous testing requirement.
Fig 2: Continuous Testing
It’s also important to engage application performance monitoring in the QA environment before running automated tests, providing feedback about any problems back to developers.
After the automated tests are executed and promotion criterion are met, the build is ready for deployment in staging or production environments.
Continuous release matches release velocity to development velocity through controlled and scalable deployments.
The goal of continuous release (sometimes referred to as continuous deployment) is to continually and automatically deploy the build into production; similar activities also can be performed in a staging environment.
As shown in Fig 3, build configuration needs to be validated, with the production environment configuration updated before being sent to production. As with application code, configuration is also maintained in the version control system and can be automated for deployment.
Fig 3: Continuous Release
After the build is deployed into production, application monitoring should be initiated to track performance and customer experience. Feedback from the production environment can be passed back to developers to guide enhancements or sprint backlogs.
This blog aims to outline core CD activities, and in subsequent articles I’ll provide additional details on entry and exit criterion, as well as metrics useful in tracking continuous development & integration, continuous testing and continuous release.
In the meantime, check out, “DevOps for Digital Leaders: Reignite Business with a Modern DevOps-Enabled Software Factory,” a new book authored by my colleagues Aruna Ravichandran, Kieran Taylor and Pete Waterhouse.
Based on learnings gained from numerous customers who have already embraced CD and DevOps, it highlights key practices for starting this journey, related ROI metrics and other useful tips.
This blog was originally posted at DevOps.com