Transient-Aware Performability of Softwarized 5G Service Chains under Non-Exponential Failure Models

This paper presents a transient-aware performability analysis of a 5G service function chain, composed of the three main nodes of the 5G core network: the Access and Mobility Management Function (AMF), the Session Management Function (SMF), and the User Plane Function (UPF). From a performance perspective, each softwarized node is modeled as an M/G/c queueing system using the Allen-Cunneen latency approximation, with service-time parameters adopted from a previously validated Open5GS-based experimental study. From an availability standpoint, each softwarized node is represented as a multi-layer virtualized stack modeled via Stochastic Reward Nets (SRNs) and composed end-to-end through Reliability Block Diagrams (RBDs). To capture non-Markovian effects arising from time-varying failure rates, exponential failure assumptions are relaxed for selected layers, and Weibull and moment-matched Gamma models are investigated under a fixed mean time to failure. Results show that, even under identical mean failure rates, different failure-time distributions yield measurable transient deviations in SFC availability, ultimately leading to nonnegligible differences in mission-time performability.