A Hardware-in-the-Loop Wireless System Design: Benchmarking NFV Network Services

5G/6G network slicing provides a method to split network infrastructure into self-contained slices, each comprising various virtual network functions (VNFs) mapped onto physical nodes. Developing robust resource allocation algorithms to efficiently embed VNFs into the available nodes of a network has been a significant focus of research into network slicing. A key component of many of these mapping algorithms is the need for accurate measurements of the resource and bandwidth requirements of the VNFs. This problem is further compounded by the performance impacts of collocating multiple VNFs on one physical node. Competition between collocated VNFs for shared hardware resources can lead to performance degradation and an inability to meet their service level agreements (SLAs). To tackle this, we propose a novel VNF profiling framework for benchmarking a set of VNFs, accounting for the effect of collocation on their performance, namely Collocation-Aware Function Mapping (CAFM). Unlike prior research, which mostly relies on simulations, we conduct comprehensive experiments on a real-world deployment environment based on the POWDER plat-form to benchmark collocated VNFs. The experimental results demonstrate that the CAFM framework can accurately detect and mitigate collocation issues causing performance degradation, ensnring that VNFscan meet their SLAs more effectively.