Advanced Domain-Driven Design for Consistency in Distributed Data-Intensive Systems

ABSTRACTMore and more data-intensive systems have emerged lately. Big Data, Artificial Intelligence, or cloud-native applications all require high scalability and availability. Data is no longer persisted in one central relational database with serialized and transactional access, but rather distributed and replicated among different nodes running only under eventual consistency. This poses a number of design challenges for software architects, as they cannot rely on a single system to mask the concurrency anomalies of concurrent access to distributed and replicated data. Based on three case studies, we developed a theory regarding how practitioners handle synchronization and consistency design challenges in distributed data-intensive applications. We also identified the "white spots" of missing design guidance needed by practitioners to handle the aforementioned challenges appropriately. We are currently evaluating our theory in the context of an action research study. In this study, we are also evaluating the novel design guidelines we are proposing in this regard, which, according to our theory, meet the needs of practitioners. Our design guidelines integrate with Domain-Driven Design, which is widely used in practice. Following the idea of multilevel serializability, we investigate the compatibility of business operations beyond commutativity. We provide concrete practical design guidance to achieve compatibility of non-commutative business operations. We also describe the basic infrastructure guarantees our design guidelines require from replication frameworks.