If I tell you the internal temperature of a specific outdoor telecom cabinet has been reported as 55⁰C (131⁰F) multiple times, would that concern you? Do you think urgent intervention is required? If it’s midsummer and the cabinet is located on the French Riviera, probably not. But if these temperature readings are collected around midnight from a cabinet at the top of the Matterhorn, certainly yes!
Data context is crucial
The issue here is context. Fundamental to your own understanding is not just having an abundance of high-quality data (although this is essential) but also knowing the context of this data – the “where”, “when” and “how” of the information. It’s fundamental to artificial intelligence (AI) systems too. Having both high volumes and full context achieves data that is actionable and trustworthy for AI analysis.
The underlying challenge
There’s an underlying challenge to this. AI and machine learning (ML) training typically takes place in cloud or data center environments; this is because training makes heavy computational and storage demands. However, by the time data arrives in the cloud or data center, its context may have been lost (the AI system can’t tell its French Riviera from its Matterhorn) and typically it is too late to rebuild this context.
This has a negative impact on data quality and prevents effective comparison and exploitation of the data, especially over time. Fortunately, there’s now an innovative solution to this challenge.
The role of customer-premises equipment (CPE)
Today’s customer-premises equipment (CPE) can run sophisticated software to facilitate data collection, using AI/ML or traditional methods, at the edge along with highly secure data transmission to the cloud or data center for training purposes. Once trained, models can be deployed at the edge, Before transmission, performing AI/ML analyses on this local, real-time data – the network edge is the optimal place for AI/ML systems to render decisions (inference), gathering all the context that’s required.
Enabled by over-the-air deployment capabilities, software-enabled CPE can become a set of pivotal assets for implementing AI/ML strategies at the network edge.
Introducing a comprehensive solution
The Nuvla.io platform, developed by SixSq and a part of Ekinops, offers an all-encompassing solution for managing CPE. This platform supports local data processing, enables secure data transfer and integrates data science tools, featuring:
-
Data lakes: Simplified management of unstructured data using cloud object storage, with a metadata catalog for easy navigation
-
Data warehouses: Efficient handling of structured data through time-series databases and metadata catalogs, making data easy to use for AI training
-
Edge data retrieval: Leverages standard equipment for data capture and processing, enabling both cloud storage and edge computing
-
Contextual integrity: Guarantees data integrity and provenance, essential for reliable AI-driven insights
-
Edge AI inference: performing AI/ML analysis at the edge, near the source of, and where decisions matter
To maximize flexibility and foster innovation, the Nuvla.io platform supports a diverse array of applications: open source, proprietary and custom. This ensures seamless integration into the ecosystem without the need for extensive app redevelopment.
Embedded in Ekinops’ next-gen CPE, Nuvla.io kickstarts a robust AI/ML strategy for businesses.
Practical application at the edge
There are many ways in which CPE-powered AI strategies can enhance edge computing capabilities.
An important example is in telecom operations. Utilizing SNMP probes, the telecom organization can collect extensive data on CPE performance and system health and use this to train AI support agents to provide proactive customer support and optimize customer service levels. Once these virtual agents are trained and deployed on CPE, there’s no further need to transmit data to the cloud. Instead, data is analyzed locally and, only when required, the operator can receive summaries and alerts.
Similarly, thinking about energy management, the telecom organization can gather essential data by measuring power consumption or inferring it through temperature metrics. It can use this to devise strategies, also with and for its customers, that reduce energy usage, underlining the value of local data processing and analysis.
Another application is utility monitoring, using the example of electric grid monitoring. Deployment of sensors connected to CPE facilitates the regular transmission of critical data, enabling advanced strategies like anomaly detection and predictive maintenance.
From structured data collection to analyzing unstructured data sourced from video and audio feeds, the possibilities are vast. With this readily accessible solution, the organization can improve product and service quality, reduce defects and malfunctions, improve energy efficiency and lower CO2 emissions, make better use of human capital and much more.
A final thought
Delivering edge AI through existing CPE investments and infrastructure is a readily accessible solution for each telecom organization. With high-quality data collection and processing at the network edge, this can unlock valuable insights, streamline operations and deliver competitive advantage. Ekinops’ CPE solutions, powered by the Nuvla.io platform, provide an orchestrated data ecosystem to foster AI success at the edge of every organization.
This post was first published on the EdgeIR website.
