1 Getting Started with AI in Cloud Scale Monetization

Learn how to get started with artificial intelligence (AI) and Machine Learning (ML) using the deployment assets provided for Oracle Communications Cloud Scale Monetization. These assets include container images, Helm charts, configuration files, and other components that assist you in installing, deploying, and managing AI services on Kubernetes.

You can use these resources in both Oracle Cloud Infrastructure (OCI) and other supported environments.

Topics in the document:

Overview of the Deployment Package

The deployment package provides all components required to deploy, manage, and monitor data, training, and prediction services for AI-based workflows on Kubernetes. The package includes:

  • Ready-to-use Docker images and Helm charts for each service (data service, training service, and prediction service), allowing you to orchestrate containers in your Kubernetes environment. Update the values.yaml file as needed to reflect your specific configuration requirements.

  • Dockerfiles and build files that enable you to apply minor updates to the base operating system or add additional dependencies.

  • Sample configuration files, Python scripts, and shell scripts that serve as references for customizing your deployment framework.

  • Sample Grafana dashboard JSON files for each service, designed to streamline performance monitoring and observability with Prometheus and Tempo.

Tasks for Preparing Your Cloud Native Environment

Prepare your system for the BRM cloud native deployment by performing the following high-level tasks:

  1. Downloading the Helm charts for the cloud native deployment. See "Downloading Packages for the Cloud Native Helm Charts and Docker Files".

  2. Downloading the cloud native images in one of these ways:

    1. From the Oracle Container Registry. To do so, see "Pulling Images from the Oracle Container Registry".

    2. From the Oracle Software Delivery website. To do so, see "Downloading Images from Oracle Software Delivery Website".

Downloading Packages for the Cloud Native Helm Charts and Docker Files

To download the packages for the cloud native helm charts and docker files, see "Downloading Packages for the BRM Cloud Native Helm Charts" in BRM Cloud Native Deployment Guide.

Table 1-1 lists the package name and archive files that need to be extracted for AI and ML services.

Table 1-1 Packages and Archive Files

Package Name Archive Files

Oracle Communications Cloud Native ML Helm Charts 16.26.0.0.0

ml-helmcharts-16.26.0.0.0.tgz

Includes:

  • data-fetch-16.26.0.tgz

  • train-utility-16.26.0.tgz

  • prediction-service-16.26.0.tgz

Oracle Communications Cloud Native ML Docker files and Artifacts Bundle 16.26.0.0.0

ml-dockerfiles-artifacts-16.26.0.0.0.tgz

Includes:

  • data_orchestrator_docker_bundle.tar.gz

  • data_processor_docker_bundle.tar.gz

  • train_orchestrator_docker_bundle.tar.gz

  • train_processor_docker_bundle.tar.gz

  • predict_orchestrator_docker_bundle.tar.gz

  • predict_service_docker_bundle.tar.gz

Oracle Communications Cloud Native BRM Billing Agent Helm Charts 16.26.0.0.0

brm-billing-agent-helmcharts-16.26.0.0.0.tgz

Includes:

  • brm-billing-agent-16.26.0.tar.gz

Oracle Communications Cloud Native BRM Billing Agent Docker files and Artifacts Bundle 16.26.0.0.0

brm-billing-agent-dockerfiles-artifacts-16.26.0.0.0.tgz

Includes:

  • brm-billing-agent-dockerfiles-artifacts.tgz

Pulling Images from the Oracle Container Registry

To pull images from the Oracle Container Registry, see "Pulling BRM Images from the Oracle Container Registry" in BRM Cloud Native Deployment Guide.

Table 1-2 lists the image names for the AI-ML cloud native components.

Table 1-2 Cloud Native Images

Component Name Image Name

ML Data Fetch Orchestrator

ml-data-fetch-orchestrator

ML Data Fetch Processor

ml-data-fetch-processor

ML Train Utility Orchestrator

ml-train-utility-orchestrator

ML Train Utility Processor

ml-train-utility-processor

ML Prediction Orchestrator

ml-prediction-orchestrator

ML Prediction Service

ml-prediction-service

BRM Billing Agent

brm-billing-agent

Downloading Images from Oracle Software Delivery Website

To download images from Oracle Software Delivery Website, see "Downloading BRM Images from the Oracle Software Delivery Website" in BRM Cloud Native Deployment Guide.

Table 1-3 lists the names of the packages and package files that you need to extract for using and configuring AI and ML services.

Table 1-3 Cloud Native Package and Package Files

Package Name Package File Name

Oracle Communications Cloud Native ML Data Service Images 16.26.0.0.0

ml-data-fetch-16.26.0.0.0.tgz

Includes:

  • ml-data-fetch-orchestrator.tar

  • ml-data-fetch-processor.tar

Oracle Communications Cloud Native ML Training Service Images 16.26.0.0.0

ml-train-utility-16.26.0.0.0.tgz

Includes:

  • ml-train-utility-orchestrator.tar

  • ml-train-utility-processor.tar

Oracle Communications Cloud Native ML Prediction Service Images 16.26.0.0.0

ml-prediction-16.26.0.0.0.tgz

Includes:

  • ml-prediction-orchestrator.tar

  • ml-prediction-service.tar

Oracle Communications Cloud Native BRM Billing Agent Image 16.26.0.0.0

brm-billing-agent-16.26.0.0.0.tgz

Includes:

  • brm-billing-agent.tar

Monitoring AI Services

You can monitor all the AI services using service metrics. These metrics allow you to track and monitor the total number of requests and the time taken by the services.

Table 1-4 lists information about the available metrics for the services.

Table 1-4 Service Metrics

Metric Name Description

aiml_data_service_request_seconds

Total time taken by the latest data service request.

aiml_data_service_request_seconds_count

Total number of data service requests received.

aiml_data_service_request_seconds_max

Maximum time taken by a data service request to process.

aiml_data_service_request_seconds_sum

Total time taken by all the data service requests to finish processing.

aiml_train_service_request_seconds

Total time taken by the latest training service request.

aiml_train_service_request_seconds_count

Total number of training service requests received.

aiml_train_service_request_seconds_max

Maximum time taken by a training service request to process.

aiml_train_service_request_seconds_sum

Total time taken by all the training service requests to finish processing.

predict_request_latency_seconds

Total time taken by the latest prediction service request.
predict_request_seconds_count

Total number of prediction service requests received.

predict_request_seconds_max

Maximum time taken by a prediction service request to process.

predict_request_seconds_sum

Total time taken by all the prediction service requests to finish processing.

In addition to these, all the metrics that are provided by Micrometer, Tempo, and Kubernetes are available.

Configuring Grafana Dashboards for AI Services

You can create dashboards in Grafana to display the metric data for your AI services.

Alternatively, you can use the sample dashboards included in the ml-helmcharts-16.26.0.0.0.tgz package.

Table 1-5 lists the sample dashboards.

Table 1-5 Sample Grafana Dashboards

File Name Description

data-fetch-grafana-dashboard.json

 Allows you to view metrics for the data service requests.

train-utility-grafana-dashboard.json

 Allows you to view metrics for the training service requests.

prediction-grafana-dashboard.json

 Allows you to view metrics for the prediction or recommendation service requests.