The Case For Change Notifications in Pull-Based Databases

Abstract

Modern web applications often require application servers to deliver updates proactively to the client. These push-based architectures, however, are notoriously hard to implement on top of existing infrastructure, because today’s databases typically only support pull-based access to data. In this paper, we first illustrate the usefulness of query change notifications and the complexity of providing them. We then describe use cases and discuss state-of-the-art systems that do provide them, before we finally propose a system architecture that offers query change notifications as an opt-in feature for existing pull-based databases. As our proposed architecture distributes computational work across a cluster of machines, we also compare scalable stream processing frameworks that could be used to implement the proposed system design.