Splitting Network Functions in 5G

5G networks are expected to support various applications with a high flexibility meeting diversity of requirements in terms of latency, data rates and massive connectivity.

A 5G network function (NF) supply a particular capability to support communication through a 5G network. NFs are normally virtualized, but some functions may need addition of more specialized hardware. NFs can be the functions that are common functions which are essential for all applications, for example, authentication and identity management NFs. On the other side, there are some functions that are not appropriate for all the use cases. For instance, a mobility management function such as handover can be used for the enhanced mobile broadband communication (eMBB) applications [1].

The new architecture of RAN, centralized RAN (C-RAN) that has recently gained momentum, decouples most Base Station (BS) functions from Remote Radio Heads (RRHs) to be pooled at Central Units (CUs) [2,3].  All RRHs are connected to CUs through the transport network. Figure 1, shows the structure of C-RAN with three RRHs which is connected via transport network to one CU. The functionalities that can be decoupled comprise channel encoding and error correction decoding, modulation/demodulation, resource mapping/demapping, channel estimation and equalization, Fast Fourier transform (FFT) and its inverse, analog-to-digital and digital-to-analog conversion, and antenna radio transmission and reception.

In this architecture, an operator, based on different application requirements is able to decide dynamically each function module to be realized in either the CU or RRHs (known as functional split). To reduce the fronthaul (FH) traffic amount, some modules can be migrated to the RRH side and other functions shifted to central unit (CU). However, the functionality at RRHs can be just as basic signal and analog processing known as Distributed RAN (D-RAN) [4].

The main advantages of C-RAN architecture are bulleted as following:

• Simplifies the structure of remote radio heads (RRHs)
• Lower operation and deployment costs
• Enables virtualization and slicing of RAN
• Allows the flexible allocation of a pool of radio and computational resources
• Controls the transitions from distributed BSs to a centralized RAN
• Migrates a hardware-defined infrastructure to a software-defined environment
• Maximizes spectrum efficiency and hardware usage.

Brief overview of different Functional Split options:

Figure 2. Different functional split options [2]

Figure 2, shows an overview of the different functional split options. The most common options are explained as following [2]:

PHY-layer split (option 6):

This split known as C-RAN, the highest centralization and coordination which enables a more efficient resource management and can be realized only with an ideal fronthaul which consumes very high bandwidth and has very low delay bounds.

MAC-layer split (option 4):

The MAC layer and the layers above it are pooled on CU with centralized scheduling (as MAC is in CU) for several RRHs. This split allows synchronized multi-cell coordination for CoMP and eICIC, but requires a low-latency fronthaul and has significant traffic overheads.

RLC-layer option (option 3):

The RLC layer and other layers above it are virtualized at the BBU. The failure over transport network may also be recovered using the end-to-end ARQ mechanism at CU. This may provide protection for critical data .This option also reduces the fronthaul latency constraints as real-time scheduling is performed locally in the RRH.

PDCP-layer option (option 2):

This option runs the PDCP functions at the BBU and may use any type of fronthaul network. The main advantage of this option is the possibility to have an aggregation of different RRH technologies (e.g. 5G, LTE, and WiFi).

Figure 3. Cloud-RAN Flexible Functionality Splitting

Figure 3, shows an example of Cloud-assisted functionality flexibility in C-RAN. As explained C-RAN is highest centralized architecture: most of processing, control, and management functionalities are migrated into the BBU pool, and the basic RF functionality remains in RRHs. However, due to various demands of applications, a fully centralized system is not optimal in all scenarios. For example, uRLLC users need more decentralized functional splits to reduce the HARQ delay. As seen in this figure, with the software-defined environment (e.g., SDN/NFV), the C-RAN operator implement a flexible functionality splitting instead of a fully centralized system which is directly related with the application requirements.

This post convey the possible splitting of network functions in 5G and describes the common options for splitting for the future radio access networks. The pros and cons of each options is also explored.

References:

[1] NGMN Alliance, “NGMN 5G White Paper,” Tech. Rep., 2015.

[2] 3GPP, “TR 38.801. Technical Specification Group Radio Access Network; Study on new radio access technology: Radio access architecture and interfaces (Release 14),” 2017.

[3] A. Garcia-Saavedra, J. X. Salvat, X. Li, and X. Costa-Perez, “WizHaul:On the Centralization Degree of Cloud RAN Next Generation Fronthaul,” IEEE Transactions on Mobile Computing, p. 1, 2018.

[4] J. Tang, R. Wen, T. Q. S. Quek and M. Peng, “Fully Exploiting Cloud Computing to Achieve a Green and Flexible C-RAN,” in IEEE Communications Magazine, vol. 55, no. 11, pp. 40-46, Nov. 2017.