5 keys to a more automated Network Operations Center (NOC)

The NOC needs processes that automate how the network identifies and resolves service-impacting incidents in real-time. Or, even better, that can prevent incidents before they happen. Reacting to negative events or customer tickets is inefficient and costly. Automation and Machine Learning can scale your ability to predict and prevent issues before they occur.

The need to consolidate and process information quickly is paramount to the success of any network operations team. Until now, Communication Service Providers (CSPs), Managed Service Providers (MSPs) and other enterprises have struggled to visualize their expanding networks quickly and accurately in a singular view, relying on legacy tools and manual practices to monitor critical network functions and services. The proliferation of inventory systems, siloed applications, and the fractured network infrastructures brought together through acquisitions, has created significant visibility gaps to the NOC, negatively impacting productivity and increasing costs.

Once you have consolidated data in one platform, you need to quickly pinpoint, analyze and resolve the root cause of service-impacting events. A system like Assure1® helps you eliminate and suppress massive amounts of noise to ensure your operations team always acts correctly against incidents that typically result in impacted services.

With ML and event analytics, you can leverage industry-standard ML algorithms with special data filters to normalize data, ensuring correct patterns are fed into the ML engine.

Using these data streams, the solution helps you detect anomalies, such as temporal deviations, statistical rarities and unusual behaviors, to generate a singular root causal event. Root causal events contain suppression patterns that filter out noise to improve NOC operators’ rate of predictability to resolve problems versus responding to a storm of event alarms (again, allowing you to be proactive instead of reactive).

At Federos, we talk a lot about actionability because it is the key to effective automation. Operations teams must shift to an actionability mindset in order to drive automation.

ML and event analytics rounds out the three-prong Assure1® strategy for providing customers with industry-leading root cause analysis (RCA). Federos delivers three types of RCA, and the final one is tied to actionability that requires a human:

• Topological RCA by leveraging physical and virtual topology discovery
• Unsupervised Machine Learning RCA that learns from patterns and does not require topology
• Supervised RCA, where operators can flag noise fields and tie them to known root causes

• Inventory Drift: Discover when inventory is drifting and automate a trouble ticket (this can happen 20, 30, 100 times a day). Assure1® Universal Topology can quickly and accurately depict topological changes in near real-time. It includes a fully integrated cross-domain topology and relationship management function to handle any technology–logical and physical.
• Event Storms and Dips: Driven by event storms (or sudden dips in events) that are caused by a singular root cause. For example: cut fiber and element management systems disconnect.
• Abnormal Behavior: Driven by learning the noise fields of every device, down to ports on switches. The abnormal behavior rule generates and escalates events based on anomalies not common to that port or device. For example, a core router port that has previously been stable but suddenly begins having issues, would be flagged and escalated for analysis.
• NOC Operational Performance: Looks at how different types of events are handled and learns how each kind of event is managed in the NOC. Based on this information, the solution sends an alert when an event is abnormally handled. For example, if a NOC operator acknowledges a downed port by adding a journal entry and then clearing the alarm, that incident would be “learned” by Assure1® as normal for that type of event. In this case, in the future if someone accidentally cleared an event without working on it, that action would raise an alarm.