The telecommunication networks of the future will not only have to support millions of 4G and 5G subscribers, but must also manage a huge number of connected internet-of-things (IoT) devices. With the need to meet exponentially growing data and signalling requirements, a new approach is needed to cope with the unpredictable and surging demands placed on modern networks.
Jio Platforms, a subsidiary of Reliance Industries, turned to machine learning to autonomously manage its large communication infrastructure. With a modest budget of $1m, Jio Platforms designed and implemented Atom, an artificial intelligence-based platform, from scratch within 12 months to process more than 500 billion records a day.
At its heart, Atom, which helped Jio Platforms clinch the telecoms category in the Computer Weekly Innovation Awards APAC, is a disaggregated data lake platform tailored to enable smarter network operations using machine learning.
Atom – an acronym for Adaptive Troubleshooting, Operations and Management – was designed to collect and process a massive volume of network-centric statistics and events. The goal was to proactively detect anomalous network patterns and facilitate root-cause analysis and resolution – before network problems even impact operations.
Jio Platforms said Atom provides code-free operational insights, data binding and correlation. Built with automated service-level agreement (SLA) management capabilities in the workflow engine, it orchestrates operational tasks between systems for organisational transparency.
It can also offer instant notifications and live data tracking from the vast amount of data collected using virtual probes and various network functions. This is made possible by a data ingestion engine designed to process billions of documents. Immediate action therefore becomes possible, as opposed to the traditional approach of only reacting to problems.
The Atom platform provides multiple ways to create reports and dashboards on the fly. Detection includes comparisons with baseline data and monitoring of operational metrics. Once a relevant condition is identified, the system analyses the data by correlating, searching for errors, or deriving the real context of the erroneous scenario.
Focus on innovation
But why did Jio Platforms begin building this first-of-its-kind system instead of relying on a suitable commercial solution? The company said it has always worked to reduce dependence on external providers and cited the cost-related advantages of developing an in-house solution that relies on software running on standard servers. Indeed, because Atom avoids the use of proprietary probes, vendor dependencies were also eliminated on that front.
Building the entire system in-house meant Jio Platforms could focus on innovation and adopt tried-and-true practices, such as developing an open solution that interoperates well with third-party systems. Atom conforms with various standards from the European Telecommunications Standards Institute and 3GPP and has the versatility to support network functions from the edge, core, on the various layers of the IP stack, and IoT applications.
Because crucial software components are developed from the ground up, the team could incorporate high-performance considerations and state-optimised designs for application resilience from the start. Jio Platforms said Atom has real-time analytics capabilities to process 50 million records every second, as well as a record capacity of over 10 trillion with support for more than 100PB of storage.
The platform has unique anomaly detection capabilities that can drill down to individual end-nodes, whether a physical server, virtual machine or containerised service, to precisely identify problematic elements within the network.
Also, the system can understand and correlate counters and logs from the radio access network (RAN) and other systems to identify the causes of failure and take corrective actions.
“Telecommunications companies operate with very large network infrastructure with large volumes of data traffic,” said the team. “Processing and analysing this data with the help of scientific algorithms, methodologies and tools is the need of the hour.”
It was with this in mind that Jio Platforms built Atom to enable actionable intelligence from network data in real time.
Continuous demand for scaling the telecom network is to be expected over the next few years as the colossal data volumes driven by 5G become a reality. More than ever, operational procedures will have to be automated to meet the ever-growing needs of modern networks and for telecommunication firms to stay relevant.