TY - GEN

T1 - From a Monolithic Big Data System to a Microservices Event-Driven Architecture

AU - Nunes Laigner, Rodrigo

AU - Kalinowski, Marcos

AU - Diniz, Pedro

AU - Barros, Leonardo

AU - Cassino, Carlos

AU - Lemos, Melissa

AU - Arruda, Darlan

AU - Lifschitz, Sérgio

AU - Zhou, Yongluan

PY - 2020

Y1 - 2020

N2 - View references (37)[Context] Data-intensive systems, a.k.a. big data systems (BDS), are software systems that handle a large volume of data in the presence of performance quality attributes, such as scalability and availability. Before the advent of big data management systems (e.g. Cassandra) and frameworks (e.g. Spark), organizations had to cope with large data volumes with custom-tailored solutions. In particular, a decade ago, Tecgraf/PUC-Rio developed a system to monitor truck fleet in real-time and proactively detect events from the positioning data received. Over the years, the system evolved into a complex and large obsolescent code base involving a costly maintenance process. [Goal] We report our experience on replacing a legacy BDS with a microservice-based event-driven system. [Method] We applied action research, investigating the reasons that motivate the adoption of a microservice-based event-driven architecture, intervening to define the new architecture, and documenting the challenges and lessons learned. [Results] We perceived that the resulting architecture enabled easier maintenance and faultisolation. However, the myriad of technologies and the complex data flow were perceived as drawbacks. Based on the challenges faced, we highlight opportunities to improve the design of big data reactive systems. [Conclusions] We believe that our experience provides helpful takeaways for practitioners modernizing systems with data-intensive requirements. © 2020 IEEE.

AB - View references (37)[Context] Data-intensive systems, a.k.a. big data systems (BDS), are software systems that handle a large volume of data in the presence of performance quality attributes, such as scalability and availability. Before the advent of big data management systems (e.g. Cassandra) and frameworks (e.g. Spark), organizations had to cope with large data volumes with custom-tailored solutions. In particular, a decade ago, Tecgraf/PUC-Rio developed a system to monitor truck fleet in real-time and proactively detect events from the positioning data received. Over the years, the system evolved into a complex and large obsolescent code base involving a costly maintenance process. [Goal] We report our experience on replacing a legacy BDS with a microservice-based event-driven system. [Method] We applied action research, investigating the reasons that motivate the adoption of a microservice-based event-driven architecture, intervening to define the new architecture, and documenting the challenges and lessons learned. [Results] We perceived that the resulting architecture enabled easier maintenance and faultisolation. However, the myriad of technologies and the complex data flow were perceived as drawbacks. Based on the challenges faced, we highlight opportunities to improve the design of big data reactive systems. [Conclusions] We believe that our experience provides helpful takeaways for practitioners modernizing systems with data-intensive requirements. © 2020 IEEE.

M3 - Article in proceedings

SP - 213

EP - 220

BT - Proceedings of 46th Euromicro Conference on Software Engineering and Advanced Applications

PB - IEEE

T2 - 46th Euromicro Conference on Software Engineering and Advanced Applications - SEAA 2020;

Y2 - 26 August 2020 through 28 August 2020

ER -