An adaptive functional split in 5G networks

Abstract

The 5G radio access network (RAN) features a partially centralized architecture, in which a subset of the network functions are deployed in a centralized unit. Centralizing these functions reduces operating costs and enables coordination techniques. However, the more functions are centralized, the more capacity is needed on the fronthaul network connecting centralized and distributed units. In addition, the required fronthaul capacity also depends on the instantaneous user traffic, which varies over time. Therefore, in order to optimize its performance, the 5G RAN should be able to dynamically adapt its centralization level to the user traffic. In this paper, we present the design of an adaptive RAN that can switch between two different centralization options at runtime. We provide design objectives and challenges, as well as measurement results from a working implementation.

A Dynamic Functional Split in 5G Radio Access Networks

“…Regarding T s , we notice that is several orders of magnitude higher than T p and that the differences between low-and high-density scenarios have been reduced with respect to T p . This can be seen via the coherence time: t s = 8.9 s and 17.5 s according to (9), respectively. Finally, we simulate the spectral efficiency achieved by the four proposed algorithms right after the calculation of their solutions to observe the performance degradation due to their convergence time.…”

“…The authors of [8] address the problem of selecting the optimal functional split when a new virtual network is added. Finally, in [9] the authors describe a real implementation that can switch between functional splits at runtime in a few milliseconds. This previous research sets the basis for our work, which is, to the best of our knowledge, the first addressing the problem of dynamically selecting the functional split.…”

“…The binary variable c g is used to indicate whether gNB g is implementing a centralized (c g = 1) or a distributed split (c g = 0) at any given time. We assume that each gNB g can change the value of c g at runtime without service disruption and with a latency lower or equal than 10 ms, as previous research shows [9]. The decision to change the split is taken by a central entity in the CU that has information from all gNBs and UEs.…”

Large- and Small-Scale Modeling of User Traffic in 5G Networks

“…In 3G and 4G networks, the distinction between large-and small-scale models is often enough to operate them efficiently. Nonetheless, one of the main features of 5G networks is flexible and dynamic management [9], which implies reacting to changes in the network to provide better user service and minimize costs [10]. Examples of this are fast reaction to network failures, in order to enable ultra reliable low-latency communication (URLLC), handling of extremely bursty traffic from massive machine type communications (mMTC), or load balancing and optimal function placement to respond to the instantaneous loads for enhanced mobile broadband (eMBB) [11].…”

Traffic Characterization of the MAC-PHY Split in 5G Networks

“…A good understanding of the fronthaul traffic generated by the MAC-PHY split is motivated by its attractive characteristics, halfway between centralization and decentralization. Indeed, MAC-PHY is the least centralized architecture that still enables coordination techniques, such as coordinated scheduling or coordinated link adaptation, which has been exploited by previous research [5]. In this work, we provide an accurate model of the traffic generated in the MAC-PHY split for a 5G RAN.…”