A US-based Communications Software and Network Solutions Provider Implements a Cloud-native ML-based Network Traffic Classifier
The solution enables end-users to prioritize traffic dynamically and ensure optimal network usage.
With its Session Border Controllers (SBC) used in the world’s largest telcos data centers, the client is a US-based provider of Communications Software and Network Solutions. SBC is the client’s special-purpose device that protects and regulates IP communications flows. The SBC market is currently witnessing technology upgrades and rolling out features rapidly.
The client had more than 100,000 devices installed in its data centers and branch offices. As a differentiator, the client wanted a network classifier that could display the bandwidth of the apps. The current solution was Deep Packet Inspection (DPI), an advanced method to examine and manage network traffic by analyzing traffic data, adding an extra level of complexity to the solution with a need to analyze the Layer-7 data. However, DPI applied at the Open Systems Interconnection’s application layer was not the most optimized and effective for achieving a network classifier solution.
Ness worked with the client’s business and R&D teams to lay out a Network traffic classification solution to reduce the device cost (not exceeding 5% of the appliance cost) and an easy-to-use interface for large MSPs, like AT&T and Comcast to configure, monitor, and analyze the network traffic.
Ness used its knowledge of SBC platform engineering and ML to recommend a cloud-based solution over DPI. The solution collected the periodically received device-level data, prepared the data for analysis, trained the model based on data, generated a prediction model, and deployed the model on devices.
By using ML, the client saved more than $900,000. The Edge solution (patent # USPTO#11140068) was able to integrate with analytics on the cloud without specific updates and dynamically influence business policies for bandwidth provisioning, security, and blocking and unblocking apps. End-users were able to prioritize traffic dynamically and ensure optimal network usage.