부산대학교 도서관

We propose a REST-based architecture to manage performability aspects of service function chains (SFCs) which represent crucial features of cloud-native 5G networks. In particular, when deploying real-time SFCs, two key aspects must be taken into consideration: performance and availability (a.k.a. performability). To ensure a certain performance level, specific network conditions must be met (we focus on the SFC end-to-end delay bound). Likewise, to guarantee availability, resilient schemes must be introduced to keep the SFC operational even in the presence of faults (we focus on the so-called five nines availability). To automate the SFC performability management, we have designed a REST-compliant framework useful to build an SFC that meets performability requirements while minimizing deployment cost. We adopt an M / G / k queueing model to deal with performance aspects, and Stochastic Reward Networks (SRN) to face availability issues of SFCs. Finally, some parameters used to build a realistic performability model have been estimated from an SFC-like testbed.