Client-centrische Replicatie voor het Gedecentralizeerde Web

Distributed systems are currently evolving from a centralized client-server architecture to decentralized, web-based architectures. Decentralized systems use replication techniques that decentralize control, but have several challenges such as resilience, interactivity, and storage overhead. In this dissertation, we address client-centric replication for the web in three distinct environments characterized by varying trust and consistency requirements: strong eventual consistency in a trusted setting, strong consistency in a Byzantine setting, and strong eventual consistency in a Byzantine setting. Each of these scenarios presents unique challenges and requirements that need to be addressed. For the first environment, strong eventual consistency in a trusted setting, we present a Conflict-free Replicated Data Type protocol that is fully state-based, yet supports fine-grained delta-merging without keeping track of individual clients. Secondly, for strong consistency in an untrusted environment, we propose a Byzantine Fault Tolerant consensus protocol with a novel way to synchronize consensus votes between the replicas, making the protocol fully leaderless. At last, we present Conflict-free Replicated Data Type protocol that supports a Byzantine environment with strong eventual consistency, without the need to keep track of individual transactions or clients. To evaluate the proposed protocols’ effectiveness, we implemented them in three separate browser-based middlewares and assessed their performance in realistic settings, including failure scenarios. This research contributes to the advancement of replication techniques in decentralized web architectures by providing robust solutions for diverse trust and consistency requirements, ultimately enhancing the resilience, interactivity, and storage efficiency of such systems.