Noise Elimination in Network AI

Noise Elimination in Network AI puts key decisions in the hand of the customer. These decisions impact how anomaly and incident noise is eliminated, and reflect what is considered operationally relevant by each network operations team.

The old saying goes “If a tree falls in a forest and no one is around to hear it, does it make a sound?”. Monitoring and observability are similar in nature. So many alerts flow across operations screens, anomalies that require attention roll off quickly, never to be seen, and never to be acted on until a customer or application team calls requesting assistance. For those network operations teams looking to automate their trouble ticketing workflow, the situation is even worse. As trouble tickets often schedule skilled operations team members, network engineers, and SRE, automation is not possible unless noise is eliminated. Ditto for automation of the network itself.

Augtera has passed the trust barrier, at scale, with some of the largest operations in the world. Noise is eliminated not by one technology, approach, or set of unmanageable rules. Noise is eliminated through a holistic approach to noise elimination which includes customer-defined policy that determines what is or is not operationally relevant. Augtera customers are realizing 100 to 1 reduction of alerts/incident notifications on individual device failures and a 10 to 1 reduction across all alerts. The net result is a 90% reduction in the time it takes to schedule a resource to investigate an incident (MTTD *), a 50%+ reduction in time to mitigation (MTTM), and a 40%+ reduction in time repair (MTTR).

This blog summarizes key aspects noise elimination and noise elimination in network AI.

Strong Signals in Network AI

The essence of Augtera’s noise reduction is to automatically mine operationally relevant “needles” from a data haystack created by billions of streamed and acquired data points per day. The Augtera platform is monitoring, observing, and extracting “needles” 24×7, constantly mining in real-time.

AI Needles result from data haystack noise elimination in AI

There are multiple ways the Augtera platform extracts strong signals from the network operations data haystack.

  • False positive and false negative elimination
  • Autocorrelation
  • Classification
  • Macro Pattern recognition
  • Gray failure detection enables the elimination of future anomalies
  • Finite state machine for control plane changes

Most of the above involve Augtera’s network-specialized AI/ML.

Policy-Based Noise Elimination in Network AI

Augtera’s mining of strong signals is industry-leading. However, we recognize that it is the customer that ultimately determines what is relevant to them. Multiple policy-based approaches include:

  • Correlation definition through Augtera Spaces
  • Notification definition through Augtera Views
  • Customer meta-data ingestion
  • Maintenance suppression
  • Duplicate incident suppression
  • Severity suppression
  • Transient anomally definition

Conclusion

Network operations teams are struggling to improve or even maintain KPIs in the face of new complexities and an avalanche of alerts. Noise elimination is critical. The Augtera Platform is proven at scale, with transformative customer results. It is the leading noise reduction platform on the market, and it is the most comprehensive, powered by network-specialized AI/ML, guided by customer-defined policy.

Read more about Noise Elimination.

* There are multiple definitions of MTTD used by network operations teams. Here we are using a broad definition which is not the time for a tool to detect an anomaly, but the time for a skilled resource to be scheduled to attend to it. As noted previously, anomalies can be detected by tools, but never noticed by a network operations team member, and never attended to.