DPI RIP: Why Telcos need a new approach to visibility

Deep packet inspection (DPI) fails with encrypted Terabit-speed traffic; Programmable Switches coupled with AI models are the way forward.

Executive Summary
Telcos operating multi-billion dollar wireline and wireless network infrastructure are under pressure to increase profitability, requiring them to optimise infrastructure efficiency, reduce customer support costs, and create upsell products. In the absence of continuous visibility into application usage patterns (e.g. streaming video, gaming, conferencing, downloads, etc.) and user experience (e.g. video freeze, game lag, conference stutter, download speed, etc.), Telcos typically operate in the blind, ceding value of their
infrastructure investments to over-the-top beneficiaries. Unfortunately, visibility relying on traditional methods of examining packet contents (aka deep packet inspection or DPI) is starting to fail as payloads get increasingly encrypted and data rates grow to Terabits-per-second and beyond. A new approach is needed.

In this whitepaper, we describe how the challenges of encryption and scale can be overcome using a combination of two emerging technologies: programmable switching and artificial intelligence. 

Rather than inspecting individual packets, programmable switches track the behaviour of each communication “flow” at millisecond-level time scales to export a “pulse” akin to a heartbeat. The “FlowPulse” of various applications
exhibit distinct characteristics, allowing trained AI engines to distinguish thousands of applications such as Netflix videos, Twitch live streams, Call-Of-Duty games, and Zoom teleconferences and to further infer user experience explicitly in terms of video freeze, game lag spikes, and conference stutters. The result: a Terabit-speed platform that gives accurate sub-second real-time visibility into encrypted traffic streams using low-cost commodity hardware.

The implications are profound - Telcos can get deep application insights across their entire network footprint in a future-proof and affordable manner to increase their profits immediately and lower costs. They can tune their networks to maximise user experience on high-value applications like gaming, increase customer satisfaction, loyalty, and retention, and create new revenues from premium services for emerging immersive applications like cloud gaming, virtual reality, and the Metaverse.

Download White Paper

Vijay Sivaraman

Vijay, Co-Founder and CEO of Canopus, is a distinguished professor at UNSW, a leader in software-defined networking, holding four patents and over 100+ research articles, and has worked at Lucent Bell-Labs. B.Tech from IIT Delhi, PhD from UCLA

Previous
Previous

Canopus Networks Secures Investment From US-Based Konvoy Ventures

Next
Next

Rethinking Speed in Telcos: A Call to Focus on User Experience