PhD defence - Rodrigo Nunes Laigner

A Scalable and Consistent Database System for Event-driven Architectures

Abstract

Cloud computing has promoted a profound transformation in how computing services are acquired, maintained, billed, and deprecated. By enabling the pay-as-you-go model, that is, the on-demand utilization of computational resources without the burden of managing hardware infrastructure, cloud computing has prompted service providers and application maintainers to rethink the design of software systems to better adhere to this computational resource rental model. In particular, we have recently witnessed an increasing adoption of event-driven and microservice architectures in the industry. These architectures promote a highly modular software design and are characterized by deploying small and independent components that communicate via asynchronous messages, sharing no state or computational resources. Such a design in the cloud ameliorates the provisioning and specialization of computational resources holistically, adapting to the needs of each component, and prevents failures from propagating outside the boundary of a component. Despite the accelerated industrial adoption and relevant evidence that data management is a major challenge in event-driven and microservice architectures, there has been a lack of thorough investigation into the state of the practice and the major challenges faced by practitioners with regard to data management in such an architectural style, which jeopardizes the advancement of data management solutions to effectively meet practitioners' needs.
To fill this gap, this dissertation starts by presenting an in-depth investigation of the state of the practice of data management in microservices, revealing several unmet data management requirements and foundational challenges that require rethinking the design and architecture of database management systems (DBMSs). To aid this development, this research proposes Online Marketplace, a novel microservice benchmark that prescribes key data management requirements that existing benchmarks fail to incorporate, including transaction processing, query processing, event processing, constraint enforcement, and data replication, allowing for properly comparing data systems and platforms. Throughout the Online Marketplace implementation and evaluation in competing systems, we found a mismatch between data management requirements and state-of-the-art data platforms for cloud applications, requiring weaving together heterogeneous systems to achieve the requirements fully.  
Finally, this dissertation answers the long-standing question of whether event-driven architectures (EDAs) are fated to eventual consistency. We propose the virtual Microservice Oriented DataBase (vMODB), a novel distributed DBMS that tackles the problem of data inconsistency in distributed and asynchronous applications by design. We propose the virtual micro service (VMS) programming model, through which vMODB leverages application semantics to enforce ACID properties. vMODB transparently unifies event and data management into a common event-driven execution framework, all without giving up the envisioned benefits of EDAs. As a result, vMODB will significantly simplify the design and implementation of highly consistent applications based on EDA. 

Assessment Committee

Professor Panagiotis Karras (Chairperson)
Associate Professor Zoi Kaoudi
Associate Professor Paris Carbone

Leader of defense: Panagiotis Karras
IT-responsible: Konstantinos Skitsas

Supervisors

Professor Yongluan Zhou

Place

The defence is conducted as a hybrid defence.

To attend the digital defence, please follow the link:

https://ucph-ku.zoom.us/j/69249386780

Instructions if you wish to attend the defence via the digital solution: Please follow the link and hereafter the instructions to download the required -client. If the -client is incompatible with your pc, smartphone etc. you can attend via an Internet browser. Log-in in due time before to allow time to install the -client.

The physical place of the defence:

HC Ørsted Institute, Auditorium 09, Universitetsparken 5, 2100 Copenhagen Ø

Ask for a copy of the thesis here: rnl@di.ku.dk