O-RAN use cases
MEC-RAN integration for low-latency use case
As described in the 5G Network Evolution on AWS
This configuration allows user traffic to be consumed locally at
the edge site without pumping traffics to the backhaul network,
which is efficiently applicable to high-bandwidth services with
regard to saving the cost of backhaul network. In addition, if you
collocate a CU, UPF, and MEC together, on the same network
functions virtualization infrastructure (NFVI) layer such as AWS Outposts, it can help for the MEC platform and application to use
a network quality status for the Radio Network Information Service
(RNIS) through the
API
exchange within the platform
Co-hosting of CU and UPF-like network functions (NFs) and MEC applications on the AWS brings the benefit of a single pane of glass for the orchestration for all network and service applications.
RIC-CU/DU operation to optimize radio resources (traffic steering and QoE optimization)
An O-RAN architecture on AWS enables ISVs (and DSPs) to dynamically interact with the radio resources, allowing them to dictate how compute and network resources are allocated to steer traffic, improve QoS, and control Quality of Experience (QoE).
As discussed earlier, the separation of RU/DU/CU provides an opportunity to steer traffic from RUs to a pool of DUs, steer control traffic from a DU to a pool of CUs, and steer traffic from RAN to a pool of 5GC resources, such as from DU to a pool of UPFs.
AWS helps you achieve traffic steering, as defined in 3GPP TR 23.793. By using a Telco data lake on AWS, operators can feed telemetry, infrastructure, and application data from their access networks (5G and 4G), from UE environments (such as Wi-Fi, 5G private networks, and so on) to predict when it becomes beneficial to offload PDU sessions from the 5G radio network to an underlying 4G network, or to a Wi-Fi network.
Amazon SageMaker
AWS enables you easily use data from a multitude of sources when combined into a data lake
Network slicing, and service level specifications (SLS) fulfillment
The idea of network slicing is to create virtualized logical networks over a physical network which consists of RANs, core networks, and transport networks. This virtual network overlaying allows the end customers, including business companies, to have an isolated and tailored network connectivity for their own business purpose.
For example, a customer can establish a network enabling low latency and ultra-reliability communications for a critical Internet of Things (IoT) use case. Another example would be the networks of connected cars. The concept of connected cars has been introduced with in-car entertainment, assisted or fully automated driving, and maintenance data gathering. On the 5G network, car manufacturers could create a virtual network with the required service level specifications to accommodate millions of connected cars.
Network slicing on the 3GPP network components like RAN has been standardized in terms of interface and functionality. In addition, one important component is needed to manage and orchestrate network slicing across RAN and core network. This management and orchestration entity oversees the operator's entire network, and creates or deletes a slice based on the customers' demands. The management and orchestration are responsible for allocating network resources and managing the life cycle of network slices.
The network slicing management and orchestration can be implemented on the AWS Cloud, as shown in the following figure. By using AWS services such as Amazon EKS (for CNF management), Amazon RDS (for data management), AWS Lambda, CodePipeline (for lifecycle and resource management), and AWS CloudFormation (for infrastructure management), the network slice manager can create a slice by allocating and configuring network resources through the AWS programmable infrastructure and well-defined APIs.
All the created slices and allocated resources can be monitored by the network slicing manager, which provides full visibility to the operator via a graphical view. In addition, AWS provides APIs to allocate resources on AWS Outposts, which are placed in the operator's own data centers or corporate edge sites so that the network slicing manager can control the on-premises resources. The following figure shows the 5G network architecture and the AWS services for network slicing.
All management operations are performed via AWS APIs, which enables network operators to have no dependency on specific resource mapping across a wide range of network domains. The consistent infrastructure APIs further allows the development of new services such as AI-powered monitoring and service assurance, by using AWS services such as Amazon SageMaker.
DevOps, CI/CD, and network management (a single pane of glass)
As described in the 5G Network
Evolution on AWS
Multiple CD pipelines are deployed per individual vendor, that are
kicked-off with the upload of updated Docker images or
configuration (such as helm charts and YAML files). Each
respective pipeline runs through various stages for vetting the
updates by deploying it first on test environments for unit
testing, later in the staging environment for system-level
integration testing, and finally in production using
blue/green,
canary-based
From an RIC perspective, it is important to manage each
characteristic of underlying network resources such as CPU,
hardware accelerators (such as
FPGA