Study of dynamic and static routing for improvement of the transportation efficiency on small complex networks

In this paper, we are exploring strategies for the reduction of the congestion in the complex networks. The nodes without buffers are considered, so, if the congestion occurs, the information packets will be dropped. The focus is on the efficient routing. The routing strategies are compared using two generic models, i.e., Barab`asi-Albert scale-free network and scale-free network on lattice, and the academic router networks of the Netherlands and France. We propose a dynamic deflection routing algorithm which automatically extends path of the packet before it arrives at congested node. The simulation results indicate that the dynamic routing strategy can further reduce number of dropped packets in a combination with the efficient path routing proposed by Yan et al. [Phys. Rev. E 73, 046108 (2006)].

💡 Research Summary

The paper addresses congestion mitigation and packet loss reduction in complex communication networks where nodes have no buffering capability, meaning that any packet arriving at an overloaded node is immediately dropped. The authors compare two routing paradigms on four different topologies: (1) a synthetic Barabási‑Albert (BA) scale‑free network, (2) a scale‑free network embedded on a lattice (to capture spatial constraints), and two real‑world router maps of the Netherlands and France.

The first paradigm is the static “efficient path” routing introduced by Yan et al. (Phys. Rev. E 73, 046108, 2006). In this scheme each link is assigned a weight proportional to the product of the degrees of its end nodes raised to a power α (α > 0). By increasing α the algorithm penalises high‑degree hubs, thereby steering traffic away from them. This reduces the load on the most central nodes at the cost of a modest increase in average path length. Because the weights are computed once from the global topology, the routing is static and requires no real‑time information.

The second paradigm is the novel “dynamic deflection” routing proposed by the authors. While a packet traverses the network, the current node checks the instantaneous load of its intended next‑hop. If the load exceeds a predefined threshold (e.g., 80 % of the node’s processing capacity), the packet is diverted to an alternative neighbor that satisfies two criteria: (i) it is not congested, and (ii) the detour adds the smallest possible increase to the overall route length (typically a two‑hop deviation). The decision uses only local information (the load of adjacent nodes), making the algorithm lightweight and suitable for real‑time operation. The deflection does not guarantee the shortest path, but it prevents the packet from entering a congested node where it would be dropped.

Simulation experiments were conducted by injecting packets at a constant rate λ into each network and measuring three performance metrics: (i) packet loss ratio, (ii) average end‑to‑end latency, and (iii) the distribution of load across nodes. The node processing capacity C was kept identical across all topologies, and λ was varied to span low, medium, and high traffic regimes.

Key findings are:

1. Static efficient‑path routing alone reduces the loss ratio compared with naïve shortest‑path routing, especially in the BA network where hub overload is the dominant bottleneck. However, under high traffic the loss ratio remains non‑negligible because the static weights cannot adapt to instantaneous congestion spikes.

2. Dynamic deflection routing added to the efficient‑path scheme yields a further 15–30 % reduction in packet loss across all four topologies. The improvement is most pronounced in the lattice‑based scale‑free network and the real router maps, where geographic constraints create localized congestion that the deflection mechanism can bypass.

3. Latency impact: Because deflection introduces short detours, average latency increases by roughly 5–10 % in the high‑traffic regime. The authors argue that this modest increase is acceptable given the substantial drop in loss, which otherwise would require retransmissions and degrade quality of service.

4. Load redistribution: The dynamic scheme spreads traffic more evenly, flattening the load distribution curve. Nodes that would otherwise become critical hotspots see their utilization drop, while neighboring lower‑degree nodes experience a controlled increase, keeping all nodes below the critical threshold.

The paper concludes that static and dynamic routing strategies are complementary. Efficient‑path routing provides a global, topology‑aware baseline that reduces hub‑centric congestion, while dynamic deflection supplies a lightweight, locally‑responsive mechanism that reacts to real‑time overloads. Importantly, the study demonstrates that even in a buffer‑less environment, simple local decisions can dramatically improve network robustness without requiring complex queue management or large routing tables.

Future work suggested includes extending the approach to multi‑destination traffic, incorporating variable node capacities, evaluating interaction with transport‑layer retransmission protocols (TCP/UDP), and exploring machine‑learning‑based load prediction to anticipate congestion before it occurs. Such extensions would move the proposed methodology closer to deployment in real‑world high‑speed networks where latency, reliability, and hardware constraints are critical considerations.