Midhaul Performance Modelling Using Commodity Hardware C-RAN Testbed

Midhauls introduce new characteristics to wireless networks. We study implementation performance behaviour to make appropriate network and algorithm design decisions soft–real-time midhaul-based radio access network implementations. A model is developed based on data obtained using a testbed, which can be used to estimate midhaul latency as a function of the number of communicating nodes. The testbed comprises one central unit (CU) controlling a variable number of distributed units (DU) synchronized over Precision Time Protocol (PTP). Average reporting latencies of 266μs\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$266~\upmu \hbox {s}$$\end{document} were observed with 16 DUs. Typical jitter performance was 99.99 % of values below 471.75μs\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$471.75~\upmu \hbox {s}$$\end{document} but maximum values up to an order of magnitude larger. Variations in performance of up to 16.2 times more deadline misses were observed between best and worst performing DUs. A model fitted to the obtained data estimates the latency and jitter of CU-DU communication as a function of the number DUs. Results indicate suitability for application with moderate synchronization requirements, for example positioning, but insufficient for the most stringent uses case such ultra-reliable communication.