1. Scope

OSM (Open Source MANO) is a well-known ETSI-hosted project providing an open source Management and Orchestration (MANO) stack aligned with ETSI NFV Information Models. The AI-Agents for OSM project is intended for automating the deployment of Artificial Intelligence Agents on the VNFs orchestrated from the OSM platform. The aim is to be able to provide the Virtual Network Functions (VNFs) deployed in OSM with artificial intelligence capabilities. The following figure illustrates the general concept:

AI-Agents - General Concept

As we see, individual AI-Agents can be attached to specific VNFs associated to different Network Services to perform AI related tasks. Of course, using AI-Agents in VNFs is optional, so there can be VNFs with or without AI-Agent.

The VNFs with AI-Agents will have additional AI/ML capabilities associated to the Service Assurance (SA) set of actions already available in OSM, namely:
  1. Alerting actions (VNF metrics can trigger threshold-violation alarms), and

  2. Auto-scaling actions (automatically increase or reduce the number of VNF instances).

Without AI-Agents for OSM these actions are typically executed based on the definition of simple rules or thresholds in OSM, associated to certain by-default metrics (CPU usage, memory usage…), but using AI-Agents for OSM this behavior can be enriched, enabling the possibility of triggering those alerting or scaling actions based also on more complex metrics leveraging AI/ML algorithms.

In fact, the AI-Agents for OSM can be used to enhance the VNFs deployed on OSM, allowing (for instance) to define proactive alerting or scaling policies based on forecasting certain service usage patterns; also, to trigger scaling actions based on image recognition or data clustering algorithms, among other AI/ML-based applications.

In summary, AI-Agents for OSM allows to go beyond the purely reactive OSM behaviour and the limited set of metrics defined on it, allowing the integration of AI/ML capabilities so expanding the by-default OSM functionality.

The deployment of the AI-Agents is based on the Execution Environments (EE) feature introduced in OSM Release EIGHT. If you want to know more details about the internal details and the implementation go to the next Architecture Overview section. If you want to go straight forward to deploy a simple ‘hello world’ example you can jump to the Usage section (you will find an example to trigger scaling actions based on image recognition algorithms).