Speed Matters: The Role of Timely Data in Service Assurance in Network Operations

Network service assurance is vital to success in the fiercely competitive telecom landscape. Service assurance implies maintaining the highest quality of service, ensuring minimum network disruption, and rapidly identifying and resolving network faults. When you have data, what you do with it is critical for telecoms. The timeliness varies, generally falling into three categories: real-time (within a second or less), fast-time (in a minute or less), and KPI-time (every 5/15 minutes or more). Each bucket has a purpose, meets specific use cases, requires different AI/ML/analytics approaches, and entails unique complexities.

Real-Time Data (One Second or Less)

Real-time data are crucial for use cases requiring immediate actions or decisions, such as automated network fault detection, traffic control, and predictive maintenance. With near-instant access to data, telecom providers can identify and resolve issues by focusing on the network itself to prevent larger network disruptions.

In this bucket, the data comes from the network nodes themselves. The AI/ML tools often involve complex event processing, real-time analytics, and machine learning models focused on anomaly detection. Deploying these models enables immediate response to potential issues, such as traffic congestion or signal failure.

The primary challenge is ensuring that the network is up and stays up and that the real-time signals in the data will flag any issues. This real-time data looks at the network rather than the services. Indeed, gathering s services and subscriber-level information at this granularity would be a big challenge. We must look to the next bucket for a fuller picture of the network, service, and subscribers.

Fast-Time Data (One Minute or Less)

Fast-time data collection is beneficial for scenarios that need rapid but not necessarily instantaneous responses, such as quality of service (QoS) management, customer experience management, and service-level agreement (SLA) compliance checks.

AI/ML applications for this data include predictive analytics and machine learning models designed to monitor and optimize network performance. Anomaly detection on subscribers is one of the fastest ways to detect issues related to subscribers or services. Fast-time data allows for quick responses to problems while leaving some room for more complex analysis compared to real-time data.

One challenge here is maintaining the balance between speed and accuracy. Processing and gathering data within a minute requires a thoughtful approach to collecting relevant data; while swift, the analysis remains comprehensive. More challenges include:

• Processing the volume of subscriber events.
• Choosing the suitable anomaly detection algorithm.
• Quickly linking those anomalies to tried-and-tested next-best actions.

KPI-Time Data (5/15 minutes or more)

KPI-time data is for more strategic, longer-term information and decisions such as network planning, load balancing, and trend identification. However, in Network Service Assurance, it is these KPIs that help monitor and troubleshoot issues on the network. These longer intervals allow for deeper analysis and provide a holistic view of network performance.

This data is where analytics can apply more sophisticated models, such as decision trees, clustering algorithms, and deep learning, for predictive modeling. These techniques help identify patterns and trends for better network and service planning and decision-making with insights into services, devices, locations, and subscriber experience.

In this case, the difficulty often resides in who uses this data and for what. When used to address real-time or fast-time use cases, the data can be stale when the investigation starts, sometimes hours after the event. This data is also not workable for real-time/fast-time closed-loop scenarios.

In Conclusion

Each data acquisition bucket serves specific purposes, served by suitable AI/ML/analytics applications and unique complexities. Recognizing these differences and applying appropriate strategies is critical to effectively assuring network, service quality, and customer experience while doing so efficiently.

The future of telecoms lies in the right data, at the right granularity, applying AI, and powering the right decisions.