Homing Policies may come from 3 different sources:
- Vendor: specify homing policies related to VNF. E.g., hardware requirements like CPU pinning, NUMA, SRIOV, etc. (VNFD)
- Service Architect: specify homing policies for a given service like vCPE. Policies may relate one or more VNFs that constitutes the service. E.g., distance thresholds between two VNFs of a service, minimizing distance between customer and VNF, etc. (NSD)
- ONAP Operator/Administrator: specify homing policies that are overarching across multiple services. E.g., given two placements that are equally performant pick the one with lowest cost, no cloud-region should have more than x% over-subscription.
Information that is required for homing is obtained at multiple levels of granularity:
- Cloud-regions and existing service (or VNF) instances from A&AI.
- Static information like cloud/cluster capabilities from A&AI (MC exports these information to A&AI)
- Dynamic runtime information like available tenant quotas, capacities etc. from MC (through DMaaP ?)
- Instantaneous and aggregated WAN capacity from SDNC