Validate the EC p-star approximation on larger network instances

Validate the complexity advantage of the Effective Congestion p-star approximation on substantially larger network instances, to determine whether its reduced per-step evaluation cost remains effective beyond the evaluated topologies.

Background

The paper introduces Effective Congestion (EC) as a deadline-aware routing metric and proposes an interface-level p-star approximation that evaluates congestion using one reference path per interface rather than separately for every interface–path pair. This reduces the per-step evaluation cost from O(|P|) to O(|E|).

The experiments on Hierarchical, Grid, and Abilene topologies show that the approximation is generally robust, although it can trail the exact path-dependent EC formulation in structurally asymmetric or tight-deadline settings. The paper explicitly leaves validation on substantially larger instances unresolved, making scalability beyond the tested networks an open direction.