Surviving the Storm: The Impacts of Open RAN Disaggregation on Latency and Resilience

TL;DR

The paper analyzes how Open RAN disaggregation (CU/DU/RU) adds signaling latency during UE attachment and elevates the risk of signaling storms, due to open interfaces like F1-C and O-FH. It builds a delay-modeling framework that decomposes $D_t$, $D_p$, $D_q$, and $D_r$, and compares monolithic RAN with Open RAN using M/M/1 queuing and derived service rates $\mu_{Monolithic}=1/\sum_{i=1}^{M}( t_{p,i} + D_{RU-BBU})$ and $\mu_{OpenRAN}=1/\sum_{i=1}^{M}( t_{p,i} + D_{RU-CU})$, highlighting higher delays for Open RAN under heavy load. A mass UE-attachment scenario demonstrates how arrival rate spikes ($\lambda$ from 20 to 200 UEs/s) cause queue buildup, with a parallel M/M/c extension describing capacity effects. The authors introduce a resilience framework with a utility function $u(t)$ and a Lyapunov-drift based adaptation that dynamically adjusts the number of servers $c(t)$ to optimize a drift-plus-penalty objective, achieving up to 286% resilience improvement and resilience scores near 0.96 under favorable conditions. The work shows that while Open RAN disaggregation can raise latency and congestion risk, its architectural flexibility enables adaptive congestion control and faster resource deployment to enhance overall system resilience against signaling storms.

Abstract

The development of Open Radio Access Networks (Open RAN), with their disaggregated architectures and virtualization of network functions, has brought considerable flexibility and cost savings to mobile networks. However, these architectural advancements introduce additional latency during the initial attachment procedure of User Equipment (UE), increasing the risk of signaling storms. This paper investigates the latency impact due to disaggregation of the Base-band Unit (BBU) into the Central Unit (CU) and Distributed Unit (DU). Specifically, we model the delays induced due to disaggregation on UE attachment, analyzing the performance under varying load conditions, and sensitivity to processing times. We demonstrate that while both monolithic and Open RAN architectures experience performance degradation under high-load conditions, Open RAN's added overheads can increase its susceptibility to congestion and signaling storms. However, Open RAN's inherent flexibility, enabled by disaggregation and virtualization, allows efficient deployment of resources, faster service deployment, and adaptive congestion control mechanisms to mitigate these risks and enhance overall system resilience. Thereby, we quantify resilience by introducing a new utility function and propose a novel adaptation mechanism to reinforce Open RAN's robustness against signaling storms. Our results show that the proposed adaptive mechanism significantly enhances resilience, achieving improvements of up to 286% over fixed configurations, with resilience scores approaching 0.96 under optimal conditions. While simulation results show that Open RAN disaggregation increases attachment latency and susceptibility to signaling congestion, they also highlight that its architectural flexibility can mitigate these effects, improving resilience under high-load conditions.

Figures (10)

• Figure 1: Open RAN versus monolithic RAN architectures.
• Figure 2: O-FH and F1-C protocol stacks.
• Figure 3: 5G NR Initial UE attachment procedure.
• Figure 4: Transmission Delay for RRC Messages with Various Transmission Rates.
• Figure 5: Impact of Utilization ($\rho$) on Average Delay per UE ($W$).