Noll fördröjning i självvy: En Implementering av Konfliktfria Replikerade Datatyper i Unity : En Prestandautvärdering av Operationsbaserade CRDTs

Interactive applications are deeply integrated in modern society, and form an integral part of everyday life. The need for low latency in these applications become more prevalent as technology standards and the scale of applications increase. This need for low latency is paramount in the field of Extended Reality (XR). Latency in XR becomes especially noticeable when a user’s own actions are delayed. This concept is known as Self-View Latency (SVL). High SVL is known to induce Virtual Reality (VR) sickness, and reduce user immersion. One way to address this problem is to introduce Conflict-free Replicated Data Types (CRDTs) as part of a local-first computing paradigm for XR applications. These data types allow the user to apply local-first updates to achieve close-to-zero SVL, while guaranteeing conflict-free synchronization of states. In this thesis, we have identified data types that can be treated as CRDTs. We have investigated what support current networking frameworks have for CRDTs, and implemented a benchmark to measure the difference in latency. We have performed the benchmark in Unity using their own networking library Netcode for GameObjects (NGO). This benchmark compares a Client-Server topology, with a custom implementation of a Distributed Authority, and a custom Relay-Server that uses one CRDT, more specifically, the vector transformation of a sphere in Three-Dimensional (3D) space. The results of the benchmark showed that it is possible, for some operations that meet the requirements of CRDTs, to gain a close-to-zero SVL for the client that issued the operation. This extreme reduction in SVL additionally benefits all clients within a session, with potential implications for improving user immersion. Additionally, our benchmark shows that by implementing CRDTs, it is possible to reduce the total network traffic of the system. However, this effect diminishes as the number of clients increase. The number of clients used for the benchmark varied between 2 and 16. Further research is required for the implications of CRDTs scalability and performance in the context of XR applications. Lastly, research on how to effectively combine existing paradigms with CRDTs to optimize latency wherever possible, could also shed light on their practical use in a server-authoritative environment.