<<place holder for documenting the design to facilitate the Acumos integration with DCAE>>
For Frankfurt release, DCAE will add new software component named "Acumos-DCAE Adapter" Adapter which will expose interface to receive new ML models into ONAP/DCAE from Acumos. which will provide a client interface to receive ML models from an Acumos catalog to ONAP DCAE. This adapter will generate the required onboarding metadata artifacts (Microservice component spec, data-format, policy-model etc) into DCAE MOD flow targeted under Self Serve Control Loops v2
Architecture
Adapter API
Sample Models
...
) and to onboard them directly into the DCAE CLI-DB. The subsequent Service Assurance flow design, Microservice configurations, and blueprint generation will be supported by the new DCAE MOD components, which will be delivered under Frankfurt Self Serve Control Loops v2
The "Acumos-DCAE Adapter" transforms an Acumos ML model into an ONAP compatible DCAE Microservice in order for it to run in an ONAP DCAE environment.
- The Adapter provides a DCAE model runner to wrap the ML model from Acumos to in order to produce a microservice that is compatible in DMaaP environment.
- The Adapter generates Microservice metadata (i.e., the Component Spec.json and Dataformat.json) for onboarding to DCAE Design environment.
- The Adapter generates k8s/Docker executable image for onboarding to DockerCentral (Nexus).
Architecture
Adapter API
- The "Acumos-DCAE Adapter" supports Acumos E5 client interface for federated ML models from an Acumos catalog.
- The "Acumos-DCAE Adapter" uses DCAE Onboarding APIs for automated mS onboarding to CLI-DB, also submits the mS Image to DockerCentral image repository.
Sample Models
Repository
Usecase Targeted
Due to lack of specific end-to-end usecase (in Frankfurt), integration was demonstrated using generic models from AT&T/Acumos Instance into DCAE-MOD.
Demo recording and details - 2020-04-16 DCAE Demo
Deliverables
- Helm Chart
- Acumos adapter docker container
Refer Acumos Adapter Installation