OCNWDAF Features

4.1 Time Series Database

The OCNWDAF uses a time series database. The benefits of using a time series database are listed below:

The database searches faster as it is organized based on time and not any other key values. It can ingest millions of records, store trillions, and maintain query latencies from a sub-second level in just a few seconds. Unlike any public cloud data warehouse, database query latency is at least three times faster.
Purging the older data is much cheaper than relational databases. Information is stored based on time in segments. An entire time segment can be quickly and easily cleared, or even dropped from the cluster but not from deep storage, still available at a higher latency for some rare access scenarios.
Another significant benefit of using this database is data roll-up. The database can roll the data into 15-minute, 1-hour, and daily entries. The database can be designed in such a manner that the minor entries can start ageing out in a few months and the 15-minute entries can start ageing out in six months. After a year, the 1-hour entries can age out. This functionality significantly reduces the amount of data that has to be stored but still provides scope for extensive historical data.

4.2 Support for AI Model

The OCNWDAF supports the industry's best ML frameworks, such as TensorFlow and scikit-learn. Many supported models, including Neural Networks and specialized versions such as Long Short Term Memory, are used.

4.3 Support for ML Model and Repository

The main constraints of machine learning models are the time necessary to train the model and the accuracy observed from the model. We are focusing on a two-prong approach described below to ensure both the constraints are minimized:

The ML models will be developed for the use cases and analytics categories will be used in a very general and broad manner. This further ensures:
- The ML models can automatically self-optimize by using the industry’s best hyper-parameter optimization algorithms.

All ML model storage will be in the ONNX format, along with an analytics type, model version, data version (define the scope of data used), and timestamp. Any external model support added should be in ONNX format.

4.4 OCNWDAF Multicloud Strategy

Oracle is a platinum member of the Cloud Native Computing Foundation (CNCF) and aligns with the CNCFs vision of an open, cloud-native, and standard approach to develop applications. Based in-depth understanding of service reliability, Oracle strives to create foundational platform services where applications are built to provide service reliability.

Oracle's OCNWDAF and Cloud Native Environment (OC-CNE) are based on industries leading open-source services and tools, this enable operations and management of OCNWDAF in a production-grade environment. Oracle has adopted software architecture principles to support Communications Service Providers multicloud strategy.

The Cloud Native Framework represents a collection of shared services which OC-NWDAF uses. Oracle maintains a reference Cloud Native Framework utilising a selection of open-source components and tools for development. The deployment CNE must provide all the Cloud Native Framework services but is free to select the specific version and component providing the service. For example, Oracle's OC-NWDAF supports operation with CSPs' version of Prometheus and can support either Jaeger or Zipkin for distributed tracing. A DevOps CI/CD model is supported to ensure OC-NWDAF is thoroughly tested and is fully operational in the deployment environment.
Kubernetes' environment provides runtime and lifecycle management services. Like the Cloud Native Framework, Oracle maintains a reference Kubernetes environment but allows CSPs to select the specific version used in the deployment CNE. OC-NWDAF runs in a containerized environment, uses Kubernetes DNS for service discovery, and can support a variety of Container Network Interfaces which integrate Oracle NFs networking requirements with the deployment SDN environment. The services lifecycle is controlled, deployment specific configuration is permitted and environment files are dynamically associated with the OC-NWDAF. Helm charts are used for packaging the component microservices into the OC-NWDAF applications. The configured repository handles the versioning available for each service and component microservices. Operators can enable Kubernetes deployments to manage the lifecycle of the OC-NWDAF, including the initial roll-out of the service, roll-out of a canary release to validate a new version, and roll-out of an upgrade to a new version.
Oracle continuously delivers new functionality into the Oracle Portal. Customers are notified and may choose to trigger their CI/CD workflow to deploy the new functionality. Oracles Communications Global Business Unit implements and executes automated unit tests for each component and service as part of the build process. Automated integration tests are run using the reference Cloud Native Framework to ensure that each component's runtime services and exposed APIs operate as expected. Other automated integration tests ensure APIs and NF components are functional and the overall business logic of the NF is operational. Oracle CGBU includes the subset of these test artefacts (which validate operation with the runtime environment and Cloud Native Framework) with the images of the NF components. Customers can incorporate these automated tests in the target deployment environment.
Oracle Communications Automated Test Tools and Scripts (ATS) helps operators to automate the complete testing lifecycle of 5G NFs. This aids in accelerating innovation as software delivery times are significantly shortened. There is a considerable benefit in deploying an automated testing solution in terms of cost, effort, and overall test coverage. With the adoption of DevOps and CI/CD in the Telecommunication domain, number of software releases has significantly increased thus making manual testing very challenging. Oracle Communications ATS helps operators to execute functional, regression and performance test cases easily without user intervention. As a result, operators can quickly deploy new software releases and rapidly roll out new features.