ML & AI Innovations Elevate Enterprise Network Performance
The demand on distributed enterprise networks is at an all-time high. More users, more devices, more applications (including an explosion of cloud-based apps) strain broadband networks and put a greater burden on IT resources. Applying machine learning (ML) and AI (artificial intelligence) helps prioritize network data, predict network traffic, and apply triage – resulting in higher-performing enterprise networks.
Data as a Valuable Asset
AI applies machine learning to solve problems: successful AI relies on ML and in turn, ML feeds ondata. Access to large quantities of data, then, is one of the most important assets in the creation of high-performing AI. As the number of users, devices and applications on enterprise networks grows, the volume of data on associated wide area networks (WAN) skyrockets.
Hughes Network Systems (HUGHES), which recently announced the first self-healing WAN solution for enterprise customers, credits the innovation in part to the large data lakes the company amasses as a managed network services provider for 20,000 distributed customer networks. In that role, Hughes accesses data from hundreds of different service providers, recording hundreds of metrics every minute from tens of thousands of sites.
Additional data is collected regularly from millions of VSAT locations worldwide. This volume of data – tens of terabytes per year – is a distinct advantage to any organization seeking to develop AI solutions to optimize application and network performance.
Applying ML for Insight
Employing ML to examine data from a variety of angles on an ongoing basis helps to identify specific metrics in locations throughout the distributed network, as well as analyze the impact those metrics may have on end users. This constant analysis via ML provides multiple insights for enterprise customers across a wide variety of metrics.
For instance, ML may identify one site within the network that consistently experiences an overuse of data. Using ML, we can see why and predict what adjustment will alleviate the problem.
In other cases, it might recognize that a particular Internet service provider in a certain geographic region experienced a spike in traffic loss, then to analyze whether the spike is normal or an anomaly. While it may take a domain expert hours, or even days, to reach a conclusion on a single metric, ML delivers near-real-time analysis. Hughes sends data packets every 10 milliseconds to its core systems, where ML algorithms prioritize the data and determine which metrics are most essential to the customer service agents or network engineers.
From Reactive to Predictive in Enterprise Network Performance
Over time, as the network recognizes data metric patterns through ML and records how each issue is resolved, the system can identify problems much faster and more accurately. With the application of AI technology, the network can automatically and proactively prescribe solutions to troubleshoot problems – much like a doctor who, based on years of medical experience, can quickly identify a patient’s illness based on their symptoms and then prescribe the appropriate remedy.
This is the foundation of the new “self-healing” feature.
The combination of ML and AI technology automatically predicts and preempts an undesirable network behavior. Deployed across 32,000 enterprise sites already, the solution has prevented service-disrupting symptoms in 70% of cases. That success rate of autonomous correction across the sites may have saved approximately 1,750 hours of network downtime in the first seven months of use. In the other 30% of cases, the system provided early diagnoses of potential hardware failure or chronic site issues so they could be addressed preemptively.
As the adage says, time is money, and reducing network downtime translates to increased productivity and healthier enterprise businesses.
While we are just starting to witness the wide-spread application of ML and AI for fast-tracked troubleshooting and autonomous corrections, these developments are only the beginning of the AI evolution in enterprise networking, and beyond.
What’s Next in Enterprise Network?
Engineers continue to work up the ML and AI solutions to the next-level network challenges. For instance, traffic prioritization is integral to high performing networks shuttling huge (and ever-increasing) amounts of data. But now that Internet traffic is widely encrypted, the data container is no longer reliably indicative of the type of traffic.
Applying ML to examine high-level characteristics of traffic may help networks properly classify and, ultimately via AI, route the traffic more efficiently. As multi-path connectivity becomes more pervasive, and more pathways are readily available (for example, the expansion of 5G or the introduction of Low Earth Orbit satellite constellations), intelligent prioritization and routing will be essential for optimizing network performance.
The telecommunications industry will continue to focus on increasing bandwidth, introducing new types of transport, and implementing multi-path solutions as the data explosion continues.
Yet, it will be ML and AI that will advance existing enterprise networks and enable future ones to reach their highest-performing potential.