A geo-replicated database is an integral part of today’s large-scale cloud services. According to the CAP theorem, it is impossible to provide both Consistency and Availability in the presence of network Partitions. On one hand, databases that guarantee strong consistency are not available under network partitions (CP databases). On the other hand, databases that adopt relaxed consistency models provide faster response and higher availability, but they expose the application to anomalous behaviors that can be challenging for programmers to handle (AP databases). In this talk, I will present AntidoteDB, an AP database that avoids most of the anomalous behaviors at the application level without sacrificing availability. We will discuss Antidote’s unique approach for improving programmability which is based on providing highly-available transactions (HATs) and Conflict-free Replicated Data Types (CRDTs). AntidoteDB has been developed within the scope of two European projects, SyncFree and LightKone.
Sara is a post-doctoral researcher at INRIA Paris. She is a member of the DELYS group which is specialized in the design and implementation of distributed systems. Her current research is focused on programming models and fault tolerance protocols for geo-replicated databases. Sara completed her PhD from the Australian National University in 2019. Her thesis explored the design of resilient programming models for high-performance computing.