Super-Netflow is the High-Octane Fuel Needed to Power the Network AI Ecosystem

NetFlow (also known as IPFIX), by which Internet routers export records of the traffic conversations they transport, has been the de facto standard for over 25 years to extract information from the network on flow endpoints, volumes, and timestamps. This standardization has fostered a multi-billion ecosystem of analytics tools that take NetFlow as a source of network data, for uses ranging from network troubleshooting and performance optimization to service assurance and security analytics. The dramatic growth of AI in recent years brings opportunities to enhance these areas, as well as open up new analytics use-cases in areas relating to automated operations, customer support, and personalised marketing, which are inadequately served by NetFlow as it stands today.

Canopus is offering Super-NetFlow – an extension that enriches NetFlow with application-level information identifying the type & provider of each traffic flow (e.g. Netflix streaming video, CS:GO gaming, Zoom conferencing, etc.), as well as its real-time quality-of-experience (QoE) measures such as video resolution, buffer stalls, gaming lag spikes, and conference stutters. In this whitepaper we show that Super-NetFlow enables new categories of AI applications that can provide better customer care using behavioral data; better capital allocation that can directly maximize user experience; and personalized marketing and advertising that can employ behavioral user segmentation. We describe how the multimedia telemetry data is represented and exported in a standardized way so it can be fed to a wide array of AI analytics platforms. We outline the schema, and illustrate it in a live deployment with representative examples of video streaming and gaming. Super-NetFlow will be the fuel that powers predictive and generative AI applications much needed by network operators to maintain and enhance profitability in the decade to come