Scalable On-the-fly Transcoding for Adaptive Streaming of Dynamic Point Clouds

TL;DR

This work introduces and evaluates a dynamic point cloud streaming system that utilizes on-the-fly transcoding and empirically shows how caching and speculative transcoding allow to significantly reduce transcoding loads, allowing to scale to a higher number of simultaneous clients.

Abstract

On-the-fly transcoding of dynamic point cloud sequences reduces storage requirements and virtually increases the number of available representations for on demand streaming scenarios. On-the-fly transcoding introduces, however, additional workload to media providers' infrastructure. While V-PCC encoded content can be efficiently transcoded by re-encoding the underlying video bitstreams, which greatly benefits from hardware-accelerated video codec implementations, the scalability of such a system remains unclear. In this work, we introduce and evaluate a dynamic point cloud streaming system that utilizes on-the-fly transcoding. We explore the limits of scalability of this system in terms of request fulfillment times, specifically evaluating the perceived user Quality of Experience. We empirically show how caching and speculative transcoding allow to significantly reduce transcoding loads, allowing to scale to a higher number of simultaneous clients.

Figures (6)

• Figure 1: The transcoding decision process: Once the server receives a segment, it either serves from storage or forwards the request to the backend, which manages the transcoding job.
• Figure 2: Experimental Setup: A central node serves as the media server. We use 1 or 2 nodes (indicated by lower opacity) for the transcoding backend and 5 client nodes to simulate a configurable number of players each. Each transcoding node in the backend is equipped with a GPU and supports 4 parallel transcoders. For player simulation, we route traffic through a software switch container, which emulates a network trace from van_der_hooft_http2-based_2016 per player.
• Figure 3: Transcoding time $t$ normalized by segment length $T$ for transcoding from rate configurations R5 to R1-R4. Averaged over 20 consecutive transcoding tasks of each looped sequence. Confidence intervals remain small.
• Figure 4: of the response time (receiving a request until server-side response). The value of the CDF at zero latency (left axis) indicates the fraction of requests fulfilled instantaneously.
• Figure 5: Average number of stall events per streaming session: Pre-encoding the lowest quality representation as a fallback (F) proves most effectively in reducing stalls for increasing simultaneous client numbers, while Caching (C) proves more effective under higher loads. Predictive Encoding (P) reduces stall events for lower number of clients, but causes faster resource exhaustion and thus more stalls when the number of clients increases.