Ensure that Helm charts are parameterized to ensure that one can instantiate the service multiple times.
Prove with some sample application to ensure that inference package can be instantiated and tested successfully.
This is breakdown of activities I see :
- Setup POC : To understand how inference works.
- Now identify the items that are to be parameterized to ensure that same Helm charts are used to deploy multiple deployments. Some in the same cluster in different namespace and some across clusters in various cloud-regions.
- Update the helm charts
- Ensure that they all work fine by deploying some test application in various deployments (this test would require its own Helm chart)
- Test above using K8S plugin API of ONAP.
- Test above using SDC and SO of ONAP
- Integrate these helm charts with rest of analytics framework
- Integrate with Prometheus helm charts.
- Integrate with Kafka broker helm charts
- Create one sample Analytics application that leverage data from collection service and use the inference using the trained model to generate alerts over Kafka bus.