Cascade Network Stability of Synchronized Traffic Load Balancing with Heterogeneous Energy Efficiency Policies

TL;DR

This work addresses cascade stability in wireless networks with heterogeneous load-balancing and sleep-mode policies, motivated by Open RAN (ORAN) architectures. It introduces a generalized dynamic model that encompasses identical and non-identical load dynamics, and derives a cascade stability criterion based on the network Laplacian spectrum, valid for arbitrary topologies. The authors analyze both homogeneous and heterogeneous cases, showing that synchronization to a desirable equilibrium $l_i=1$ hinges on negative derivatives of local dynamics and coupling terms; sleep-induced heterogeneity can destabilize the synchronized state. An ORAN SMO architecture is proposed to monitor policies via the Central Unit (CU) and enforce stability decisions, with simulations on random networks and PPP-based topologies illustrating convergence behavior and the impact of sleep thresholds on stability. The results offer a theoretically grounded framework for designing and validating load-balancing policies in large-scale, heterogeneous networks, with practical implications for energy efficiency and QoS in 5G/ORAN deployments.

Abstract

Cascade stability of load balancing is critical for ensuring high efficiency service delivery and preventing undesirable handovers. In energy efficient networks that employ diverse sleep mode operations, handing over traffic to neighbouring cells' expanded coverage must be done with minimal side effects. Current research is largely concerned with designing distributed and centralized efficient load balancing policies that are locally stable. There is a major research gap in identifying large-scale cascade stability for networks with heterogeneous load balancing policies arising from diverse plug-and-play sleep mode policies in ORAN, which will cause heterogeneity in the network stability behaviour. Here, we investigate whether cells arbitrarily connected for load balancing and having an arbitrary number undergoing sleep mode can: (i) synchronize to a desirable load-balancing state, and (ii) maintain stability. For the first time, we establish the criterion for stability and prove its validity for any general load dynamics and random network topology. Whilst its general form allows all load balancing and sleep mode dynamics to be incorporated, we propose an ORAN architecture where the network service management and orchestration (SMO) must monitor new load balancing policies to ensure overall network cascade stability.

Figures (6)

• Figure 1: (top) Open RAN architecture for stable load balancing: 1. BSs report their load balancing behavioral dynamics $f(.), g(.)$ to CU. 2. CU computes the cascade stability impact on whole network. 3. CU reports outcome to DU. 4. DU decides how to allow load balancing between RUs. (bottom) User-cell handover reality using equation \ref{['equ: handover']} and stochastic geometry model in paper.
• Figure 2: (top) Real cellular traffic demand data and load spike data. (bottom-left) Base station performs a range of baseline power down, sleep mode, and dynamic cell zooming 6502480, trading off power consumption against traffic load. (bottom-right) Equivalent load rate of change against traffic load plot. The load dynamic plot data fitted with polynomial ordinary differential equation (ODE) to enable checking of cascade stability.
• Figure 3: Examples of potential dynamics: (a) The local load dynamics $l_i$ is a BS in active mode. If $l_i<1$, this underloaded BS should attract load to increase load. If $l_i>1$, then this overloaded BS should offload to other BSs. $l_i=1$ is a desirable equilibrium for maximum service efficiency. (b) The associated offloading dynamics $g(\cdot)$ of BS $i$ in active mode. (c) The local dynamics of BS $i$ in sleep mode. If $l_i$ smaller than the threshold of sleep, BS is sleeping and offloads to others. (d) is the associated offloading dynamics of BS in sleep mode. When the BS is sleeping, even $l_i<l_j$, BS $i$ still offload to other BSs. When $l_i$ is larger than the threshold, it is active and the offloading dynamics is similar to the BSs in active mode.
• Figure 4: (a) is the eigenvalue distribution of random network 1. (b) is the eigenvalue distribution of random network 2. (c) is the synchronization process of load balancing related to network 1. (d) is the synchronization process of load balancing related to network 2.
• Figure 5: (a) The random geographic BSs generated by PPP, where blue nodes are BSs, grey lines are the boundary of BSs. (b) Complex network of BSs corresponding to (a), where black lines are offloading relations. (c) Eigenvalue distribution of the network. (d) Load balancing synchronization process.