A Quantitative Approach to Optimize 5GC Refactoring for Minimum Signaling Latency and Resource Allocation

This article proposes a quantitative approach to optimizing the 5G core (5GC) network refactoring as an example. Our previous study formulated the refactoring problem to minimize queuing delay and resource allocation cost directly in the M/M/k model and utilized the optimization tool, GUROBI, to derive an optimal refactored 5GC architecture, abbreviated as GUR-5GC. However, it is time-consuming; this approach for refactoring optimization is not feasible for dynamic scaling design. We design two quantitative indicators, message exchange reduction (MER) and merging utilization degradation (MUD), to evaluate the impacts of merging certain network functions. Moreover, the problem of calculating the two quantitative indicators can be reduced to a string-matching problem. Then, we reconstructed the optimization model formulation by using MER and MUD indicators in the objective functions instead of the queuing delay and resource allocation cost, since the optimization tool (CPLEX) Mathematical Programming (MP) and Constraint Programming (CP) models could not solve the original problem. Then, we utilized the CPLEX MP Model optimizer integrated with Pareto optimality to derive the CPLEX-based Refactored 5GC (CPR-5GC). In addition, we use a CURE-based (clustering) algorithm with MER and MUD by performing the quantitative analysis to derive the CURE-based Refactored 5GC (CUR-5GC) architecture. Finally, we analyzed the performance of the 5GC, GUR-5GC, CPR-5GC, and CUR-5GC. Moreover, we evaluate them in terms of queuing delay and scaling side effects. The performance results show that CPR-5GC and CUR-5GC outperform the original 5GC and are close to the GUR-5GC; the two heuristic algorithms for 5GC refactoring are feasible and practical.