Artificial intelligence has two important influences on the modern network: new classes of applications and…
AWS, SD-WAN and Beyond: 4 Breakthrough Use Cases of Machine Learning
What do Google Translate, Spotify and modern SD-WAN solutions have in common? While they couldn’t seem more different from one another on the surface, all of them are benefiting from ongoing improvements in machine learning algorithms. Thanks to ML, these platforms are better at:
- Gleaning insights from large repositories of data, with minimal supervision from humans: This learning technique helped Google Translate improve the quality of its Chinese-English translations, a notoriously difficult task for software.
- Making proactive decisions: Spotify, like Netflix and other on-demand streaming services, deliver recommendations based on listening habits, in effect attempting to solve a “problem” (namely what to consume next) before it even occurs.
- Automating previously manual workflows: SD-WAN vendors are looking to increase the role of ML in their solutions, for better anticipation of network congestion, more accurate security monitoring and advanced path selection for priority traffic.
The overall market for ML solutions is expected to expand at a 43.6 percent compound annual growth rate from 2018 to 2022, according to BCC Research. The potential use cases are vast, but here are four to keep an eye on in the networking domain in particular:
1. ML-capable Cameras in Amazon Web Services
Many organization need to ensure reliable connections between their WANs and SaaS/IaaS applications running in Amazon Web Services (AWS). SD-WAN leaders like Talari Networks deliver safe connectivity to clouds like AWS, via secure Internet breakouts, and can also aggregate broadband and Amazon Direct Connect links. These features in turn enhance the value of the growing portfolio of ML services in AWS.
For example, the AWS DeepLens is a specialized camera with machine learning functionality, specifically for video and photo analysis. It can be set up as an edge device that makes signed calls to AWS. Its powerful hardware and reliable network connectivity make it ideal for extensive ML-powered image processing.
2. Fault Prediction in SD-WAN Platforms
Today’s SD-WANs are miles ahead of traditional ones in terms of automating tasks and adjusting in real-time to evolving network conditions. ML algorithms could add another dimension to SD-WAN’s capabilities, by boosting its fault prediction functionality.
In practice, this enhancement would resemble predictive maintenance in manufacturing, which estimates when systems will require maintenance and, by doing so, often extends the useful life of these assets through timely repairs. SD-WAN platforms could similarly draw upon AI and ML to predict faults, make preemptive adjustments and notify administrators. Accordingly, they could help resolve major issues before end users even notice they’re happening.
3. Superior Network Path Selection
The rise of SaaS, IaaS and Internet of Things (IoT) applications has placed significant new demands on WAN architectures unequipped to meet them. SD-WAN offers a way out, through more liquid bandwidth that’s aggregated from multiple link types and centrally managed with full visibility. Its dynamic path selection, based on constant measurement of indicators such as latency and packet loss, also supports service assurance.
With machine learning functionality, SD-WANs could refine their congestion avoidance feature sets. More specifically, proper ML training might enable the controller to better anticipate adverse conditions on a particular link and automatically divert its traffic to a more stable one.
4. Improved Differentiation of VoIP and Non-VoIP Traffic
Machine learning algorithms have been tested as alternatives to the typical port-based and transport-layer classification methodologies for classifying network traffic types. A 2016 research report demonstrated significant improvements on this front via an ML approach.
Its researchers applied ML to differentiate VoIP and non-VoIP traffic, noting the importance of accurate identification to enforcing Quality of Service and intrusion detection. ML-powered classification systems could boost SD-WANs, which already excel in supporting highly demanding apps like VoIP.
A Talari failsafe SD-WAN is a forward-looking solution for your network in the era of machine learning. Click below to learn more or request a demo today.