Application Configuration Repository for Adaptive Service-Based Systems: Overcoming Challenges in an Evolutionary Online Advertising Environment

Software engineering has greatly evolved in recent years. Today applications are deployed on heterogeneous distributed infra-structure from mobile devices to cloud computing. Service-oriented architectures, such as SOA and REST Web Services, have been widely used to efficiently design high-availability, scalable and reliable systems for dynamic business environments based on a distributed infra-structure. Despite the improvements these architectures have made to enhance the evolvability of systems, there are some challenges that still need to be overcome. More concretely, service-based systems and development teams are constantly under pressure from business stakeholders who continuously increase their demands for changes in systems. This paper describes a configuration-based approach that can empower adaptive mechanisms in order to overcome this challenge. It presents a solution based on a centralized application configuration repository service specially designed as a RESTful web service API to provide the benefits of configuration, such as adaptability, to high-availability, scalable and loosely coupled systems, allowing them to respond quickly to changes. The solution was successfully implemented in an evolutionary online advertising system used by the largest Brazilian web-portal, responsible for processing 5 billion ad requests per month. It allowed the design of a self-adaptive advertisement ranking mechanism that continuously evolves the system configuration, without human supervision. The adoption of this solution was responsible for a drastic increase in the amount of changes applied in this advertising environment. It also greatly reduced the time from conceiving a new change to having it working in the system. Moreover, the solution is available as open source and it has also being used by several other service-based systems.