In my last blog, we briefly touched upon the advanced use cases that are possible in 5G. One such use case is Network Slicing. 5G cellular services should broadly support use cases including enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communications (URLLC) and Massive Internet of Things (MIoT). On the lines of these use cases, 5G is designed to serve different verticals such as industrial (robotics), transportation (automated vehicles), media (streaming), entertainment (gaming) and many others. There is also another interesting possibility of creating multiple differentiated logical networks within an operator network for MVNOs. All of these allow operators to conceive and monetize new types of services.
One challenge though is the ability of the 5G network to cater to the varied requirements of these use cases in an end-to-end fashion. How does the 5G network handle this challenge? We will discuss the framework that enables network slicing, the O-RAN Alliance’s approach towards this feature, and the further evolution of network slicing.
Network slicing – The 3GPP view
Conceptually a network slice is a logical network that is created over the underlying physical network to serve a specific use case or service in an end-to-end fashion. The 3GPP-defined slice types are eMBB, URLLC, MIoT and V2X (vehicle to everything communication). The concept of a slice applies not just to the network, but to the user device or terminal as well. A 5G terminal can support multiple slices simultaneously.
The logical slices can be created in the radio access network, core network and transport network to support end-to-end logical networks. For example, this could mean specific instances of network functions in the control and user plane for dedicated processing for network slices. New network functions were defined in 5G core network for network slicing related call flows and slice administration. Network functions that support slicing are termed “slice aware.”
In the radio access network, a RAN slice (part of the network slice) is created to provide appropriate resources in the RAN for the service requirements. There could be dedicated or shared RAN slices based on the use cases. A slice-aware scheduler in the Distributed Unit (DU) allocates radio resources as per slice requirements. One of the key challenges in serving multiple slices is to develop a sophisticated scheduling algorithm in the DU to meet competing and conflicting needs of different use cases like eMBB and URLLC among others.
In the Centralized Unit (CU), slicing might require differentiated handling of data traffic. Often, separate CU User Plane (CU-UP) instances are created to serve the slices in a dedicated fashion.
In view of the challenges in realizing the RAN slices, the continuous monitoring of slice performance is often required. Appropriate KPIs of slice performance can be reported back to an orchestrator or a RAN controller (or a slice manager) for needed corrective actions in a closed loop manner. This is an active area of research. Radisys partnered with KDDI Research to collaborate on one such study to demonstrate RAN slicing and its control via a RAN controller for KPI based analytics.
Slicing and orchestration
Network slices are created dynamically for serving specific use cases and decommissioned when they are no longer needed. Given the dynamic nature of slice creation, activation and deactivation, there is a need for automating these operations. This is often achieved by using various levels of orchestrators in the mobile network. 3GPP specifications address in great detail, management and orchestration of network slices. Open source projects like ONAP (Open Network Automation Platform) incorporated slice management aspects defined by 3GPP. With the availability of orchestrators with such features, automated and sophisticated network slice management is now possible. There is a significant amount of work being done to provide automation features and user-friendly interfaces to design and deploy network slices.
In the radio access network, slice management and orchestration can be realized in different ways. How do we implement network slicing in open and disaggregated networks? Let’s look at the O-RAN Alliance’s approach.
O-RAN architecture and network slicing
With its disaggregated CU and DU nodes, the O-RAN architecture naturally fits the network slicing need for scaling, isolation and dedicated processing, etc. The O-RAN Alliance’s approach to RAN slicing is to employ SMO (Service Management and Orchestration) and RIC (RAN Intelligent Controller) to create and manage RAN slices. The use of artificial intelligence and machine learning in non-real-time RIC enables slice performance analytics and optimization. The near-real-time RIC can host slice aware xApps and algorithms and collect KPIs over theE2 interface to assure slice Service Level Agreement (SLA) and dynamically control RAN nodes to achieve slice optimization. A simplified view of this slicing architecture is presented in the diagram below.
The O-RAN Alliance Working Group 1 (WG1) is currently focused on use cases such as RAN slice SLA assurance and multi-vendor slices. To achieve the use cases, O-RAN interfaces O1, O2, A1 and E2 are undergoing further standardization. O-RAN WG1 is also working on possible deployment options of slice management functions in relation to SMOs.
Industry adoption and further evolution
The promise and potential of network slicing is very high, but adoption is still in its very early stages. The need to define fine-grained control of complex end-to-end use cases is in itself a challenge. In the RAN network, for network slicing to be successful, RAN virtualization is one of the key requirements. RAN virtualization in 5G is a work in progress for many products in deployment. And finally, the aspect of end-to-end slice orchestration and management is still evolving in terms availability of features.
The standardization of network slicing continues in 3GPP releases 16 and 17 to address further enhancements and automation. This includes the concept of Generic Network Slice Template and its parameters, and additional KPIs for each slice. With the ongoing work in 3GPP and the O-RAN Alliance, there is going to be a lot more sophistication and automation possible in future RAN networks.
- 3GPP TS 28.530 5G; Management and orchestration; Concepts, use cases and requirements
- 3GPP TS 28.533 5G; Management and orchestration; Architecture Framework
- 3GPP TR 23.400-40: Study on enhancement of network slicing; Phase 2
- O-RAN Working Group 1 Slicing Architecture v01.00