The 5G-NORMA approach

Diomidis S. Michalopoulos                                          Mark Doll                              
Nokia Bell Labs, Germany                                         Nokia Bell Labs, Germany        
diomidis.michalopoulos(at)nokia-bell-labs.com             mark.doll(at)nokia-bell-labs.com          

Peter Rost
Nokia Bell Labs, Germany
peter.m.rost@nokia-bell-labs.comy

5G-NORMA is a 5G PPP Phase 1 project, which has developed the concept of a flexible 5G architecture. This architecture fosters the support of diverse services with different requirements within a common network infrastructure.

The principal concept behind this approach is the transition from a network of physical, hardware-based entities to a network of functions. This essentially means that the fundamental network operations are decoupled from the hardware elements: In conjunction with the network function virtualization concept, any network function is not bound to a given physical device as it traditionally was, but can rather be executed in any network element with processing availability.

The first step towards this envisioned flexible architecture is the decomposition of network functions. With reference to the RAN protocol stack, this implies that some network functionalities carried out in the RAN (radio access network) are detached from the protocol layers, as they traditionally are, and can thus be executed in cloud entities such as the edge cloud. This provides an increased level of flexibility of deployment that adapts to diverse services, because instead of introducing diverse parametrizations of a common virtual network function (VNF), different VNFs can be allocated to different services.

RAN slicing

Network function decomposition is conceptually in line with the notion of RAN slicing, i.e., the virtual partitioning of the RAN into logical networks. In fact, RAN slicing represents an attractive architectural approach particularly for implementing scenarios where services with diverse requirements co-exist in a limited geographical space, such as an industrial campus.

Figure 1: The three RAN slicing options considered

Depending on the extent of the shared infrastructure across different RAN slices, we distinguish three major slicing options, as depicted in figure 1. Specifically, the upper part of figure 1, included in the green frame, corresponds to the RAN part which is not shared between slices, whereas the lower part, included in the orange frame, illustrates the shared RAN part. Following the approach developed in the 5G-NORMA project, the slice-specific RAN part is controlled by a dedicated Software Defined Mobile Network Controller (SDM-C), whereas the shared part is controlled by the SDMC coordinator (SDM-X), which spans multiple slices. In all three slicing options, the co-existence of two slices is considered, corresponding to Mobile Network Operator (MNO) 1 and 2, respectively. Such MNOs carry out different services, corresponding for example to Mobile Broadband (MBB) and Internet-of-things (IoT), respectively.

The RAN slicing option 1 corresponds to a slice-specific RAN. That is, the protocol layer functionalities are separate for the slices involved, with the only exception of the lower part of the Physical layer (i.e., the physical transponder) which is shared. The main advantage of this option is its flexibility, since each slice can be customized across the entire protocol stack. However, this increases the complexity of implementation. The typical example of this slicing option is the common deployment of legacy technologies (such as 4G) and future technologies (such as 5G).

In RAN slicing option 2, the common RAN part, includes the entire PHY and MAC layers at the data plane as well as the RRC layer at the control plane. This allows for some level of flexibility, as slice-specific QoS prioritization is utilized, while reducing the overall complexity due to the multiplexing of the common part, providing a less complex alternative to option 1. Finally, RAN slicing option 3 corresponds to a slice-aware shared RAN, where the entire RAN deployment is shared by the involved MNOs.

RAN adaptation to slices

A major aspect of the flexible architecture developed in 5G-NORMA is the ability of network functions to be dynamically configured both in terms of functional operation and physical location. In fact, 5G-NORMA reveals a new dimension in network architecture where functions are extracted from a pool of resources and are used exactly when and where they are needed. This implies that different slices are allocated different functions or function blocks. Two fundamental levels of flexible function allocation are distinguished: function selection and function placement, as illustrated in figure 2.

Depending on the slice requirements on throughput, latency, and reliability, a different set of network functions is used. For instance, for meeting the high demand of the broadband slice in terms of throughput, special network functionalities are deployed which are associated with the concept of multi-connectivity. As shown in the left part of figure 2, the PDCP layer is enhanced with an additional functionality (PDCP Split Bearer) which splits the information flow into two logical channels that are then combined at the UE. Similarly, the MAC layer can also be partially enhanced with the carrier aggregation functionality, depicted with MAC CA in figure 2, for further increase of the downlink throughput. Moreover, for facilitating multi-connectivity, the PDCP Split Bearer functionality is chosen to be placed in the edge cloud, whereas lower RAN functionalities are carried out at the antenna site in order to satisfy their real-time operation requirements.

Contrary to the broadband slice, the low latency and the mission critical slice are associated with requirements relevant to machine-type communications. Specifically, the low latency slice (c.f. figure 2, middle column) is usually associated with applications with small, bursty packets, where fast delivery is important, contrary to the achieved data rate. To this end, functions such as the PDCP Split Bearer and MAC CA are not included here, while the RLC layer is simplified to support only the unacknowledged mode of operation (RLC AM), accounting for a faster packet transmission. For the same reason, all network functionalities are placed at the antenna site, leaving the edge in a “transparent” mode of operation. As far as the mission-critical slice is concerned (c.f. figure 2, right column), the reliability level is of paramount importance. Consequently, an additional functionality is included in the PDCP layer, which is used for duplicating data across different antenna sites, exploiting macro diversity and thereby increasing the reliability. Similar to the broadband slice, such functionality is placed at the edge cloud, allowing for a centralized coordination of duplicated packets.

Figure 2: RAN slice adaptation – Function selection and placement

Conclusion

Flexible architecture designs facilitate the support of diverse services under a common infrastructure. The 5G-NORMA project develops an architecture framework that allows the RAN to be customized to the supported network slices. This is achieved by flexibly adapting the operation of network functions as well as the location where such functions are carried out.

Further information on the 5G-NORMA website –
https://5g-ppp.eu/5g-norma