Deterministic Scheduling of Periodic Messages for Low Latency in Cloud RAN (1801.07029v5)

Abstract: Cloud-RAN (C-RAN) is a cellular network architecture where processing units, previously attached to antennas, are centralized in data centers. The main challenge in meeting protocol time constraints is minimizing the latency of periodic messages exchanged between antennas and processing units. We demonstrate that statistical multiplexing introduces significant logical latency due to buffering at network nodes to prevent collisions. To address this, we propose a deterministic scheme for periodic message transmission without collisions, eliminating latency caused by buffering. We develop several algorithms to compute such schemes for star-routed networks, a common topology where all antennas share a single link. First, we show that deterministic transmission is possible without buffering when routes are short or network load is low. Under high load, we allow buffering at processing units and introduce the Periodic Minimal Latency Scheduling (PMLS) algorithm, adapted from classical scheduling methods. Experimental results indicate that even at full load, PMLS finds deterministic transmission schemes with negligible logical latency, whereas statistical multiplexing incurs substantial delays. Moreover, PMLS runs in polynomial time and scales efficiently to hundreds of antennas. Building on this approach, we also derive low-latency periodic transmission schemes that coexist with additional random network traffic. This article extends previous work presented at ICT.