Low-Latency AR Edge Cloud

Edge cloud promises to support low-latency applications by bringing computing close to the user. As a distributed computing infrastructure, it poses challenges in managing heterogeneous bandwidth and computing resources, and providing seamless service mobility. Future AR/VR applications will heavily rely on edge cloud to fulfill their intensive computing requirements in real-time and will require techniques such as application specific routing. We experimented with the low-latency AR-based smart navigation and meeting application using the ORBIT testbed. The low-latency (16โ€“47 ms) is achieved through utilizing higher bandwidth (912 Mbps), deploying closer edge cloud server (single hop), avoiding involved computing at the user equipment (OpenCV and path finding), and passing minimum processed information to the client (8โ€“1,500 Bytes). A future COSMOS experiment will rely on this preliminary work, where the inter-edge cloud bandwidth, system load, and topology, will be configured. The system will be injected with emulated service requests along with live AR user traffic. For a mix of AR applications, essential edge cloud evaluation metrics (e.g., application delay, service migration cost, and system capacity) will be obtained. It can provide performance guidelines for future edge cloud deployments.

Figure 1. Setup of the low-latency AR edge cloud experiment, where the mesh topology can be dynamically configured by COSMOS' optical network.