Eventual Consistency (EC) models are being increasingly adopted today with the compelling demand on high availability even at the expense of consistency. However, most EC systems are currently running in the wild, assuming that servers can only crash/recover and clients are benign. A recently designed secure EC system based on Byzantine Fault Tolerance (BFT) opened a long discussion with an expert, in both areas, and lead to the conclusion that "There are no BFT fans anymore…". Having spent some time on designing both models, I will try to open a discussion on the tension between both BFT and EC communities to explain its potential causes and how to mitigate some of the "claimed" caveats of both models in our secure EC proposal. The talk also aims to interact with the audience to identify the reasons that hamper the adoption of BFT in industry and what is missing in both models to come up with more reliable EC systems.
I am currently a researcher at INESC TEC (HASLab unit) and Invited Assistant Professor at the Engineering School of the University of Minho, Portugal. I hold a PhD in Informatics and Telecommunications from the University of Toulouse, France. I’ve spent six months at EPFL working with Rachid Guerraoui who also co-supervised my PhD thesis on Adaptive Byzantine Fault Tolerance. I’ve also spent one year as a postdoc at INSA de Lyon university, France, working on secure anonymous P2P communication. Currently, my research focus is on AP systems, Data Management, Infrastructure Security and Fault Tolerance, and more recently on Edge Computing; however, I am always interested in understanding, building, and improving Distributed Systems. I am currently leading INESC TEC’s contribution within the Edge Computing LightKone H2020 project. I teach Distributed Computing with Carlos Baquero at the University of Minho; and I am (co-) supervising three PhD students working on Conflict-free Replicated DataTypes (CRDTs), secure Eventual Consistency, and exactly-once delivery in "almost infinite scale" applications.
