A QoS Approach for Detecting and Managing a Fault Alarm Storm

When a large communication network encounters abnormal situations, such as multiple cable cuts or a failure of interface protocol leading to large numbers of propagating alarms may potentially cause many symptom events in multiple network domains. As a result, a network management system may be overwhelmed with a heavy alarm load scenario (a.k.a alarm storm). Appropriate fault management tools are required to provide auto-monitoring capabilities which indicate the health and status of the system and auto-alerting capabilities, which alert to conditions that impact system performance. This paper provides a method and strategy for managing a fault alarm storm by detecting unusual system performance behavior and implementing a reliably handling mechanism to maintain adequate system application performance. The paper explores a QoS approach to calculate inter-process latency based on dynamic states of various types of alarms and behaviors of system performance. It provides quality of service trending projection functionality of inter-process latency to manage CPU and memory capacity and to trigger the control mechanism in order to protect system resources. Based on this approach, it enables more efficient/accurate system performance measurement algorithms to handle the most import alarms need to be taken care of at the most critical situations.