Gathering solutions and providing APIs for their orchestration to implement continuous software delivery

Abstract

In traditional IT environments, it is common for software updates and new releases to take up to several weeks or even months to be eventually available to end users. Therefore, many IT vendors and providers of software products and services face the challenge of delivering updates considerably more frequently. This is because users, customers, and other stakeholders expect accelerated feedback loops and significantly faster responses to changing demands and issues that arise. Thus, taking this challenge seriously is of utmost economic importance for IT organizations if they wish to remain competitive. Continuous software delivery is an emerging paradigm adopted by an increasing number of organizations in order to address this challenge. It aims to drastically shorten release cycles while ensuring the delivery of high-quality software. Adopting continuous delivery essentially means to make it economical to constantly deliver changes in small batches. Infrequent high-risk releases with lots of accumulated changes are thereby replaced by a continuous stream of small and low-risk updates. To gain from the benefits of continuous delivery, a high degree of automation is required. This is technically achieved by implementing continuous delivery pipelines consisting of different application-specific stages (build, test, production, etc.) to automate most parts of the application delivery process. Each stage relies on a corresponding application environment such as a build environment or production environment. This work presents concepts and approaches to implement continuous delivery pipelines based on systematically gathered solutions to be used and orchestrated as building blocks of application environments. Initially, the presented Gather'n'Deliver method is centered around a shared knowledge base to provide the foundation for gathering, utilizing, and orchestrating diverse solutions such as deployment scripts, configuration definitions, and Cloud services. Several classification dimensions and taxonomies are discussed in order to facilitate a systematic categorization of solutions, in addition to expressing application environment requirements that are satisfied by those solutions. The presented GatherBase framework enables the collaborative and automated gathering of solutions through solution repositories. These repositories are the foundation for building diverse knowledge base variants that provide fine-grained query mechanisms to find and retrieve solutions, for example, to be used as building blocks of specific application environments. Combining and integrating diverse solutions at runtime is achieved by orchestrating their APIs. Since some solutions such as lower-level executable artifacts (deployment scripts, configuration definitions, etc.) do not immediately provide their functionality through APIs, additional APIs need to be supplied. This issue is addressed by different approaches, such as the presented Any2API framework that is intended to generate individual APIs for such artifacts. An integrated architecture in conjunction with corresponding prototype implementations aims to demonstrate the technical feasibility of the presented approaches. Finally, various validation scenarios evaluate the approaches within the scope of continuous delivery and application environments and even beyond.