Alarm Propagation Based on Topological Relationships for Telco Use Cases

In Telco environments using IBM Cloud Pak for AIOps, network infrastructure such as routers, switches, and access nodes is often modeled topologically as parent entities containing various dependent resources like physical interfaces, cards, and modules—typically represented through PartOf relationships.

Today, alarm correlation is driven primarily by matchToken logic, which maps events to topology nodes using fields like Node and SourceId. However, this does not take into account the hierarchical or structural relationships in the topology, resulting in missed or fragmented correlations—especially in cases where parent resources fail, but child components remain silent or unalarmed.

Is necessary than one alarm impact in more than one resource, that is part of the parent resource.

Current Limitation:

Device-to-Device Link:

Node A --> Interface 0/2/14 <- connectTo -> Interface 0/2/14 <- Node B

1. Node A fails → raises alarm
2. Interface 0/2/14 on Node B sees link down → raises alarm.
3. No alarm from Node A's interface 0/2/14 → no correlation occurs when resource group only have Interface 0/2/14 on A  -> Interface 0/1/13 on B 

As a result, no correlation is made between the failure of Node A and the interface alarms on Node B, because there's no direct alarm to bridge the correlation logic.

Shelf or Card Failure:

1. A shelf/card goes down in Node C.
2. All ports or services relying on that hardware lose connectivity.
3. No direct alarms from those ports → downstream effects not correlated.

Slot-Based Architecture:

1. A slot with multiple cards/modules fails.
2. Interfaces hosted on those modules don’t raise alarms individually.
3. NOC receives fragmented symptom alarms only.

Several additional scenarios exist.

Proposed Enhancement:

Introduce a flexible, configurable alarm propagation policy based on topological and structural relationships that allows:

Upstream events (e.g., at a node, shelf, or card level) to automatically propagate alarms to child components such as:

Interfaces // Ports // Cards // Modules

Configuration of edge types (PartOf, RunsOn, ConnectedTo, Contains, etc.) for propagation.

Rule-based control:

Configurable fields that the propagation should apply:

Example: 

summary  one of : Chassis Down | Ping Down | Card Failure | Hardware Failure 

and 

severity > minor

Which resource types can receive propagation: 

Example:  card | module | interface