Performance Assessment of Multi-Class 5G Chains: A Non-Product-Form Queueing Networks Approach

This work presents a performance assessment of 5G Service Function Chains (SFCs) by examining and comparing two architectural models. The first is the Mono chain model, which relies on a single path for data processing through a series of 5G nodes, ensuring straightforward and streamlined service delivery. The second is the Poly (or sliced) chain model, which leverages multiple paths for data flow, enhancing load balancing and resource distribution across nodes to improve network resilience. To evaluate the performance of these models, we introduce a performance indicator that captures two critical stages: the time required for user registration to the 5G infrastructure and the time needed for Protocol Data Unit (PDU) session establishment. From a performance standpoint, these stages are deemed crucial by the European Telecommunications Standards Institute (ETSI), as they can adversely affect both objective and subjective network parameters. Using a non-product-form queueing network approach, we develop an algorithm named ChainPerfEval, which accurately estimates the proposed performance indicator. This approach outperforms standard queueing network models, where the exponential assumption of inter-arrival and/or service times may lead to an inaccurate estimation of the performance indicator. An extensive experimental campaign is conducted using an Open5GS testbed to simulate real-world traffic scenarios, categorizing 5G flows into three priority classes: gold (high priority), silver (moderate priority), and bronze (low priority). The results provide significant insights into the trade-offs between the Mono and Poly chain models, particularly in terms of resource allocation strategies and their impact on SFC performance. Ultimately, this comprehensive analysis offers valuable and actionable recommendations for network operators seeking to optimize service delivery in multi-class 5G environments, ensuring enhanced user experience and efficient resource utilization.