CRDTs Enable Seamless Dynamic Software Updates for Distributed Apps

Researchers Seppe Wyns, Jim Bauwens, and Elisa Gonzalez Boix have introduced a novel approach to dynamic software updates in distributed systems using Conflict-free Replicated Data Types (CRDTs). The core idea is to model application updates as a new App CRDT that stores application code along with a semantic version, which defines a total order of code updates. This CRDT works in tandem with an API-compatible message delivery middleware, enabling applications to continue functioning even when some components have been updated and others have not—a critical requirement for handling backwards-incompatible changes in real-world distributed deployments.

The team implemented their approach in two distinct programming environments: AmbientTalk, an ambient-oriented language designed for distributed systems, and LuAT, a Lua-based framework for ambient computing. Their results show that the CRDT can be integrated with existing applications requiring minimal changes to the codebase. By generalizing the method across languages, the researchers demonstrate that using CRDTs for replicating code can be a universal solution for live updates. This work has implications for any distributed system that needs to evolve without downtime, from cloud services to edge computing networks.