Network Slicing Architecture and the Non-RT RIC
Truminds Software Systems
Global Presence - San Diego | Greater Boston | Gurgaon | Hyderabad | Bangalore
End-to-end network level Architecture
It is well known that the concept of network enables running multiple logical, customized networks on a shared common infrastructure . These logical, customized networks need to comply with agreed SLAs for various?vertical industry customers (or tenants) and requested functionalities. To achieve this goal, network slicing needs to be designed from an end-to-end perspective, spanning over diverse technical domains (e.g., device, access network, core network, transport network and the network management system). The figure below?captures the logical structure of this architecture.
End to end network and service orchestration
This cloud based management layer provides a unified, end-to-end view of the system as well as a seamless way to manage the system that encompasses technologies across diverse domains. This layer also includes important functions such as AI/ML based analytics, network slice life cycle management, portal for user interaction?etc.
Domain Resource Orchestration
Domain Resource Orchestration has a responsibility to provide life-cycle management (e.g creation and deletion), scale, and control?functions for compute, storage and network resources per RAN, Transport Network, and Core Network domain. For example, the RAN orchestration controller can create and delete virtual network functions (say,?virtual O-CU-CP or virtual O-DU) as needed.
Within this layer, the Non-RT RIC has been highlighted in the figure above?because this component interacts with the O-RAN components (i.e.: GNB/?DU/?CU etc.).
Network Slices and Network Slice Subnets
As indicated in the figure above, an end-to-end network slice defined by 3GPP spans across RAN, Transport and Core domain specific network slice subnets. Interestingly, the transport slice subnet is different from the other two in a significant?detail: The 3GPP Network Resource Model (NRM) includes the modeling of the NG-RAN slice subnet and the 5G core slice subnet. However,?from a transport perspective, only?the modelling of the transport network endpoints is covered by the model. It?does not?include the modelling for the 5G transport network itself, nor the modelling of?access networks other than RAN. There are two main reasons for this:
Non-RT RIC O1, and A1 interfaces
The figure below provides a contextual representation of the Service and Management orchestration framework, the non-RT RIC and the O-RAN components.
Non-RT RIC
The fundamental role?of the Non-RT RIC in O-RAN slicing architecture is to gather long term slice related data through interaction with the SMO framework and apply AI/ML based approaches interworking with the Near-RT RIC to provide?innovative RAN slicing use cases. For this purpose, Non-RT RIC should be aware of the configured RAN slice subnets and their respective SLAs. It gets this information from the SMO. In addition, Non-RT RIC may retrieve enrichment information from 3rd?party applications?enabling advanced RAN slicing technology to be applied in the O-RAN framework.
领英推荐
Thus, utilizing the configuration information, enrichment information and slice specific performance metrics received from the O-RAN Nodes, Non-RT RIC monitors long-term trends and patterns regarding RAN slice subnets' performance, and trains AI/ML models to be deployed at Near-RT RIC. The nature of communication from the Non-RT RIC towards the O-RAN consists of the following:
O1 Interface
This is the interface between O-RAN managed elements and the SMO. It is used to configure slice specific parameters of O-RAN nodes based on the service requirements of the slice. One example?is?the set of RRM (Radio Resource Management) policy attributes to provide the ratio of PRB (Physical Resource Block) allocation across slices.?O1 will also be used to configure and gather slice specific performance metrics and slice specific faults from O-RAN nodes.
Finally, as mentioned above, this interface is also used to deploy AI/ML models from the Non-RT RIC to the Near-RT RIC.
A1 interface
The Non-RT RIC guides the Near-RT RIC using “policy objectives” communicated over the A1 interface. The Near-RT RIC enables optimized RAN actions through execution of deployed AI/ML models or other slice control/slice SLA assurance xApps in near-real-time by considering both O1 configuration (e.g. static RRM policies) and received A1 policies, as well as the separate slice specific measurements retrieved by the near-RT RIC from the nodes (i.e. O-CU, O-DU etc).
In summary, the Non-RT RIC influences the behavior of the RAN in the following ways:
1.?????? Coarse grain control is directly exercised over the O1 interface by configuring various RAN parameters (e.g PRB allocation ratios across RAN slices)
2.?????? Fine grain control is indirectly exercised by communicating the policy objectives to the Near-RT RIC (e.g.: Max downlink throughput for a specific QOS class within a specific cell).
References
1.?????? O-RAN Working Group 1 Slicing Architecture (www.o-ran.org/specifications )
2.?????? Samsung Technical White Paper: “Network Slicing”
Link to RAN slicing - Why is it needed and how can it be leveraged? (Part 1) https://www.dhirubhai.net/pulse/ran-slicing-why-needed-how-can-leveraged/?trackingId=5SIa5cWXKQ%2BPVTqcn49G1w%3D%3D
Articulated by Venkatesh Potdar, AVP Engineering at Truminds.