Big data machine learning is not only disrupting edge delivery, edge compute and edge security but also the network. The traditional methods used by CDNs to deliver packets between locals is likely nearing its end, being replaced by methods that work, think and predict at the packet level. Startups like Deepfield and Jolata are introducing new ways to monitor network congestion and performance at the packet level, leveraging big data and machine learning to do things that haven’t been possible before.
Within the data center, Cisco’s Tetration, a data center telemetry platform analyzes packet level detail at line rate, then applies unsupervised machine learning to provide in-depth analytics on flow volume, application latency, dependencies, and much more.
Deepfield, acquired by Nokia is a pioneer in the field of big data machine learning network telemetry. At its core, the Deepfield platform is able to ingest petabytes of data elements in real time from hundreds of sources including BGP, DNS, syslogs, SNMP, NetFlow, and so on, and provide an in-depth real time view of the global network. Eventually, data center and network telemetry will make its way to the content delivery industry because by their very nature, CDN’s are mini-telco’s who manage complex global networks. Big data and machine learning is disrupting every layer of the content delivery network from caching to distributed storage to compute to edge security to the global network.
Being able to monitor, identify and fix problems associated with network congestion, packet loss, performance and latency in real time is becoming a reality. Better yet, self-healing networks are in sight, where an intelligent big data network leveraging supervised and unsupervised learning will be able to predict packet loss, congestion and latency before it happens, introducing a business model we yet have to see. That leads to the question, how will machine learning impact routing in general? We’ll soon find out.
