Network Service Descriptor (NSD)

The NSD references one or more VNFDs that define the individual software components of the service. Each VNFD specifies the virtual machine or container resources, software images, and monitoring parameters for a single network function. The NSD links these atomic units together, defining how many instances are required and their specific roles within the overall service topology.

Virtual Links define the logical connectivity between VNFs and Physical Network Functions (PNFs). An NSD specifies the connection points on each function and the VLs that wire them together, creating a service graph. These descriptors include bandwidth requirements, Quality of Service (QoS) parameters, and latency constraints to ensure the inter-function communication meets the service's performance profile.

The VNF Forwarding Graph defines the ordered sequence of network functions that traffic must traverse. It specifies the ingress and egress points and the chaining logic, such as:

• Linear chains: Traffic flows through functions in a strict A→B→C order.
• Branched paths: Traffic is steered to different functions based on classification rules. This graph is the blueprint for the service's data plane, dictating how packets are processed.

A single NSD can contain multiple deployment flavours to support different scaling and redundancy models for the same logical service. Each flavour defines a specific set of resource requirements and placement policies. For example:

• A 'gold' flavour might specify geographically redundant VNF instances with strict anti-affinity rules.
• A 'bronze' flavour might use a single instance with best-effort resources. This allows the same descriptor to be used for both high-availability core services and cost-sensitive edge deployments.

The NSD embeds the rules for automated operations, known as Day 2 management. These policies dictate how the orchestrator should react to events without human intervention. Key policies include:

• Scaling rules: Thresholds for auto-scaling a VNF based on CPU or session load.
• Healing rules: Actions to take upon VNF failure, such as restarting or redeploying on a new host.
• Update procedures: The allowed method for upgrading a VNF, such as a rolling update or a blue-green deployment.

Service Deployment Points are the external interfaces that expose the composed service to the outside world. The NSD defines these access points, which include public IP addresses, DNS names, and authentication parameters. SDPs abstract the internal complexity of the forwarding graph, presenting a single, consumable endpoint for client applications or other services to connect to the newly deployed network service.