Network Fault Prediction and Degradation Detection

Typical outcomes for this kind of engagement. Results depend on your data and starting point.

A network operations centre sees thousands of alarms a day, most of them noise and many of them repeats of one underlying fault. Degradation that has not yet crossed a threshold raises nothing at all, so the first signal is often a customer complaint. Engineers spend their time correlating alarms by hand to find which ones share a cause.

• Thousands of alarms a day, most of them noise or repeats
• Degradation below a threshold raising nothing until a customer calls
• Engineers correlating alarms by hand to find a shared cause
• Fixed thresholds that do not reflect how a given cell normally behaves
• Performance counters, alarms and probe data held and viewed separately

CloudGate brings alarms, performance counters and probe measurements into one time-aligned model, correlates alarms that share a root cause into a single incident, and scores cells and links for deterioration against their own normal behaviour rather than a fixed threshold. What is degrading is surfaced with the evidence behind it, and raised as one incident rather than four hundred alarms.

• Alarms, counters and probe measurements time-aligned in one model
• Alarms sharing a root cause correlated into a single incident
• Deterioration scored against each element's own normal behaviour
• Degradation surfaced before it crosses a hard threshold
• The evidence behind each prediction shown, not just a score
• Incidents raised into the existing operations workflow