XRFlux: Virtual Reality Benchmark for Edge Caching Systems

TL;DR

XRFlux introduces a Unity-based, open-source benchmark for evaluating edge caching in VR streaming by simulating multiuser VR workloads with dual FoVs (immediate and predicted) and logging object visibility to generate realistic demand traces. The framework replays these traces through an edge cache and renders content at the edge with adjustable LoD, enabling analysis of latency, QoE, and cache efficiency under various caching policies and network conditions. Key contributions include a configurable simulation model with principals and groupies, an edge- cache architecture (LRU and beyond), and a portable implementation that supports remote cloud interaction and multi-resolution content. This benchmark enables researchers and practitioners to evaluate VR edge-cloud deployments and compare caching strategies without relying on human-subject experiments, with planned extensions to more complex scenes and network protocols.

Abstract

We introduce a Unity based benchmark XRFlux for evaluating Virtual Reality (VR) delivery systems using edge-cloud caching. As VR applications and systems progress, the need to meet strict latency and Quality of Experience (QoE) requirements is increasingly evident. In the context of VR, traditional cloud architectures (e.g., remote AWS S3 for content delivery) often struggle to meet these demands, especially for users of the same application in different locations. With edge computing, resources are brought closer to users in efforts to reduce latency and improve QoEs. However, VR's dynamic nature, with changing fields of view (FoVs) and user synchronization requirements, creates various challenges for edge caching. We address the lack of suitable benchmarks and propose a framework that simulates multiuser VR scenarios while logging users' interaction with objects within their actual and predicted FoVs. The benchmark's activity log can then be played back through an edge cache to assess the resulting QoEs. This tool fills a gap by supporting research in the optimization of edge caching (and other edge-cloud functions) for VR streaming.

Figures (8)

• Figure 1: Illustration of the different components of the benchmark.
• Figure 2: Simple illustrative example of different FoVs per user; where S is depth distance for the immediate FoV, and D is the depth distance for the predicted FoV.
• Figure 3: In-game snapshot of the virtual environment from a user's FoV, while in motion. The floating objects in the scene are obtained from ShapeNet ShapeNet2015.
• Figure 4: Hit rate of shared cache to all users at different depths (cutoffs) of the FoV at 110°.
• Figure 5: Hit rate of shared cache to all users at three different cache access behaviors. The blue line with circle marker only considers the immediate FoV with angle of 110 degrees and depth $S=10$. The green line with square markers considers the immediate FoV with 140 angle degrees and depth $S=20$. The red line with triangle markers considers both FoVs, immediate and predicted, with objects being pre-fetched at the cache when they are at the predicted FoV. Lastly, the purple line with diamond markers represent the IRM results.