Model-based Data Synchronization Framework for Data-driven Serverless Applications

Abstract

Serverless Computing is a new approach to cloud computing in which the cloud provider dynamically manages the allocation of machine resources. The recent trend in writing serverless applications mainly addresses the web as well as other data-driven and eventdriven distributed applications which frees developers from maintaining a server. Developers just focus on the application logic in developing applications. Data management is the foundation of high-quality data-driven serverless applications in use today. Such data-driven applications are used on multiple devices, over a variety of connections and it is essential to offer a consistent user experience across different connection types. Particularly, the growing number of these serverless applications that allow activities such as messaging, commenting and collaboration present users with updated information using real time ability. Consequently, design and implementation of behavior of such serverless applications is very complex especially when data is shared among multiple users. To target this issue, this research introduces UMSDA (Unified Modeling Language Profile for Serverless Data-driven Applications) which adapts the concept of UML Class Diagram, Object Diagram and State Machine Diagram to model the frontend and backend requirements for data-driven serverless applications at high abstraction level. To resolve the complexity of application behavior, backend requirements containing data store and sync concepts are modelled in UMSDA. For integration of data with the view layer, UMSDA covers the frontend requirements including the user interface and data binding. As a part of research, a complete transformation engine is developed using Model-to-Text approach to automatically generate frontend and backend low level implementations of Angular 2 and GraphQL respectively from high level source UMSDA models. Finally, the validation of this research work is presented through two case studies: 1) Real-time Chat Application and 2) Weather Forecast Application, deployed on AWS (Amazon Web Services) Serverless platform. The outcomes prove that the proposed framework allows the modeling of both frontend as well as backend requirements of datadriven serverless applications with simplicity. Furthermore, a transformation engine is capable to automatically generate the deployable Angular 2 and GraphQL code. Finally, it has been concluded that the proposed framework greatly simplifies the design and implementation complexity of data-driven serverless applications to achieve certain business objectives like productivity and time to market.