Fronthaul for a Flexible Centralization in Cloud Radio Access Networks

Cloud radio access networks promise considerable benefits compared to decentralized network architectures. By centralizing baseband processing they enable smaller radio access points as well as cooperative signal processing and easier upgrades and maintenance. By further executing the processing not on dedicated hardware but on dynamically allocatable general‐purpose processors, cloud‐based networks also enable load balancing between processing elements to enhance energy and cost efficiency. However, centralization also imposes challenging requirements on the fronthaul network in terms of latency and data rate. This is especially critical if a heterogeneous fronthaul is considered, consisting not only of dedicated fibre but also of, for example, millimeter‐wave links. A flexible centralization can relax these requirements by adaptively assigning different parts of the processing chain either to the centralized baseband processors or the base stations based on the load situation, user scenario and availability of fronthaul links. This not only reduces the requirements in terms of latency and data rate but also couples the data rate to the actual user traffic. For example, if the subcarriers are demapped at the radio access points, only actually utilized resource blocks need to be forwarded. Thereby, statistical multiplexing gains can be observed that further lower the required data rate in aggregation networks. In this chapter, we provide a comprehensive overview of different functional split options and analyse their specific requirements in terms of latency and data rate. We compare these requirements to available fronthaul technologies to show the feasibility of heterogeneous fronthaul and we discuss the convergence of fronthaul and backhaul technologies. By evaluating the aggregated fronthaul traffic, we show the benefits of flexible centralization and give guidelines on how to set up the fronthaul network to avoid over‐ or under‐dimensioning.