Résumé
Programming distributed systems with replicated state
Distributed systems replicate state across machines to enable high availability, fault tolerance, low latency, and scalability. Yet we cannot have the cake and eat it: the CAP theorem demonstrates a fundamental trade-off between consistency and availability when network partitions occur. As a result, many systems adopt weak consistency models, relaxing the consistency guarantees to provide high availability and low latency. However, concurrent updates may lead to conflicts that must be resolved to ensure state convergence. Replicated Data Types (RDTs) have emerged as a principled answer to this challenge, offering convergence guarantees out of the box.
In this talk, we take a closer look at what it means for developers to build and utilise RDTs in their applications. Building highly available distributed applications with RDTs remains challenging: developers face recurring obstacles, including a lack of appropriate abstractions for replica discovery and update propagation, composing RDTs, and security. We will discuss recent results aimed at improving the adoption of RDTs, covering three fronts: 1) principled approaches to efficient RDT implementation, 2) abstractions for application-specific RDTs, and 3) verification of RDTs in a (semautomatic way to give developers confidence in their designs. We will close with an outlook on promising research directions.
Biographie
I obtained an Engineering in Informatics degree (Enginyeria en Informática) at the Universitat Politècnica de Catalunya (UPC) in 2004. After working as software developer for IBM Toronto Labs in collaboration with UPC, I started a PhD in Sciences at the Programming Technology Lab (now known as the SOFT lab) of the VUB. In 2012, I obtained my PhD in Sciences at the VUB on programming language abstractions and tools for handling partial failures in distributed applications running on mobile ad hoc networks. My PhD heavily relied on reflection and meta-level programming.
As from October 2014, I am a full-time professor at the SOFT lab leading the DisCo research group. Disco counts with 2 post-doctoral researchers and 4 PhD students focusing on programming languages and tooling for concurrent and distributed systems. Together we explore several research topics including debuggers, secure programming technology, and programming abstractions for distributed systems like replicated data types.
Besides research, I am also interested in teaching computer science to youngsters. As a former Erasmus and UPC student, I also actively contribute to the Erasmus exchange between VUB and UPC. In my personal life, running, reading, my family and Catalonia are my main hobbies.