Skip to content

Service KPI Metric Based Scaling of CNF/KNF using Prometheus exporters

Proposers

  • Rahul Kumar (Tata Elxsi)
  • Deepika E (Tata Elxsi)
  • Jayaprakash G (Tata Elxsi)
  • Ramesh Ramanathan (Tata Elxsi)

Description

This feature aims to extend the existing scaling capabilities in OSM to support CNFs and KNFs. By leveraging real-time service KPI data exposed through the node_exporter running within CNF/KNF pods, OSM can implement:

  • Dynamic CNF/KNF Scaling: OSM will dynamically scale the number of CNF/KNF instances based on predefined thresholds for specific service KPIs. This allows for automatic scaling up or down based on demand, ensuring optimal resource utilization.
  • Improved Performance: By scaling based on service KPIs, OSM can ensure the service meets performance expectations during peak loads or when resource utilization is low.

We propose this feature to enable auto-scaling of CNFs/KNFs based on the service KPI metrics obtained from the KDU using node exporter. This feature leverages existing capabilities within OSM (from Rel-14 / feature-10945), to provide a comprehensive solution for optimizing resource utilization and service performance within a cloud-native environment.

Technical Background

  • From Rel-14 using feature-10945 of OSM (#10945 (closed)) Users can add an 'exporter-endpoint' in VNFD and use the Prometheus exporters to scrape the KPI metrics directly from the VDU. Based on those metrics, autoscaling is also possible in OSM for "VNF" based NS instances.
  • Feature-11016 (#11016 (closed)) extends this capability to OSM with NGSA.

Demo or definition of done

A Robot test must be created or updated, containing at least the following steps:

  • Onboard VNF package with VNF exporter endpoints and template.
  • Instantiate the NS.
  • Verify Service KPI is flowing to OSM-Prometheus.
  • Verify the scaling based on the predefined threshold.