Reducing Deployment Latency and Improving Runtime Stability in AR/VR Platforms via Unified Services

Authors

  • Adrian K. Lau Department of Computer Science, The University of Hong Kong, Pokfulam, Hong Kong SAR Author
  • Emilia D. Fraser Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5S 2E4, Canada Author
  • Joris M. van Leeuwen Faculty of Electrical Engineering, Mathematics and Computer Science, Delft University of Technology, 2628 CD Delft, The Netherlands Author

DOI:

https://doi.org/10.71465/mrcis149

Keywords:

AR/VR platforms, service convergence, edge computing, deployment latency, motion-to-photon, telemetry, event format

Abstract

AR/VR platforms run many services across devices and networks, which can slow rollouts and reduce runtime  quality.  We  built  and  tested  a  service  convergence  approach  that  unifies device-facing and content services, and places them across edge and cloud with simple rules tied to latency targets. The study covered three regions, six device classes, and a  12-week window using a blocked cross-over schedule. We measured end-to-end timing with OpenTelemetry spans and motion-to-photon (MTP) with an optical rig. Median deployment latency fell from 128 s to 69 s (−46. 1%); p95 service-call latency fell from 214 ms to  132 ms; MTP p99 improved from 58 ms to 44 ms.  Rollback  events  during upgrades  dropped by  31%. A  shared event  format reduced duplicate logs by 41%, cut median time-to-detect from 66 s to 46 s, and lowered distinct incident clusters by 33%. These results show that treating scene, input, and telemetry as services—and placing them near users when needed—improves speed and stability and simplifies operations. The approach suits multi-device deployments; limits include three regions, six device types, and a 12-week study period.

Downloads

Published

2025-12-01