AI Use Cases
Analytics
Analytics is the systematic computational analysis of data or statistics. It is used to make data meaningful and easy to understand. AI-driven analytics is analytics that is powered by AI to accomplish the same goals as analytics, but with no explicit programming. In addition, AI-driven analytics may be able to find patterns or trends in data that developers would not have been able to code explicitly because they did not know what they were looking for in the first place.
Analytic insights gained through AI can be used in numerous ways: active monitoring, evaluating, or taking predictive actions. Models that take predictive actions identify patterns in real-time data that have occurred before a past failure, indicating a similar failure may be coming soon.
Predictive Analytics
Predictive analytics is a form of advanced data analytics that uses ML and AI to analyze real-time network data, compare this data with historical data, report on the patterns identified, and recommend actions. In turn, network administrators can maintain optimal network performance and reliability through proactive and intelligent management.
Predictive analytics starts with a baseline that is based on the data collected during normal operation. These baselines learn and adapt as the network environments change and the number of devices, users, and applications evolve. After defining “normal,” you can compare real-time operation data with that baseline to see if anything unexpected or abnormal is occurring within your network. With this streamlined approach to analyzing network data in place, you can also perform AI-driven root cause analysis on complex issues within your network and identify the most critical issues. Administrators can use this collection of information, debugging assistance, and remediation suggestions to organize the work that must be done to keep a network operating within defined parameters.
Performing maintenance based on AI insights before a failure occurs is called predictive maintenance. For example, by monitoring sensor data or the internal temperature of an appliance, an AI model could detect a degrading component before it fails and recommend this component be repaired or replaced. Addressing issues before they reach a critical status always produces a better outcome, usually avoiding downtime and likely extending the lifespan of hardware.
The ability to analyze vast amounts of data and make intelligent decisions saves companies time and money. Through proper implementation of AI-driven predictive analytics, organizations can see a real return of value in the way of:
Reduced network downtime
Improved cost efficiency
Increased transparency on how a network is operating
Better insights on usage statistics
AI-Driven Analytics Process
In practical terms, implementing AI-driven analytics involves several key steps. First, data must be collected from various network devices and sensors. This data includes performance metrics, usage patterns, and environmental conditions. Data that has not been purposed yet or incoming real-time data must be transformed into a more meaningful format that aligns with a given use case. Data cleaning ensures consistency and accuracy in a dataset. AI models trained on faulty data will derive flawed patterns and make incorrect inferences. If you train a model on a dataset containing numerous faults and failures labeled as normal operations, the model will not be able to distinguish the patterns that indicate those failures from normal data.
Data segregation, which is the process of sorting data into various categories based on specific attributes or criteria, is another important consideration. Data segregation can help with keeping track of data and reduce the overhead involved in using that data in a meaningful way in the future. The opposite of segregation is aggregation, a process in which data from various sources is combined into one collection. An aggregated dataset can be used as the training dataset for ML models that glean more profound insights into the system than would be readily apparent from a single data point or source. ML models recognize patterns in the data and begin to develop a baseline that includes the following:
How the data appears under regular operation
How data appears leading up to, during, and after a failure
How the data appears during training of the data
Once the data is processed, it is typically stored and should be used to continue improving the training dataset for future versions of the ML model. It is rare for a system to stay the same. Predictably, new features will lead to more users or usage, more requests, more CPU cycles, and the need to scale. ML models evaluate the incoming real-time data to detect patterns that could indicate that action should be taken. Once patterns are identified, these systems can present the information to network administrators through a data visualization dashboard or other means.
Network Optimization
In today’s technological landscape, the scope and complexity of networks are ever-changing and growing. Modern networking often includes route optimization and analysis software, like Cisco Cisco Secure Network Analytics, AppDynamics, and ThousandEyes, to name a few. These technologies help to improve how a network handles traffic. Dynamic routing dramatically simplifies the process of calculating complex routes. This approach accounts for factors such as current network traffic, capacity, availability, and other important considerations. Integrating AI into the equation further aids in optimizing bandwidth and improving the end-user experience. In this topic, you will look at how AI can help optimize network traffic and avoid congestion, resource contention, and service outages. You will also explore how Meraki, a Cisco cloud-based network management platform, uses AI to manage networking operations.
Suppose that a network segment is congested due to heavy traffic loads. Using AI, traffic can be directed through other paths to avoid congestion, ensuring that bottlenecks are prevented, and that high performance is always maintained. AI can guarantee uninterrupted service provisioning for essential applications under different network capacities by utilizing real-time information and sophisticated algorithms prioritizing critical functions.
AI helps avoid resource contention and keeps components within optimum working conditions through continuous network monitoring and adjusting resource allocations according to need. This kind of foresight informs strategic investments in infrastructure to support anticipated future growth without affecting present performance levels.
What sets AI apart from traditional traffic management solutions is its ability to look deeper at historical network data collected across regions, networks, and use cases. This capability enables an AI-driven solution to not only optimize routes based on the way things are now, but also for future occurrences such as holidays, product releases, and other events. When coupled with network management solutions, this combination has the potential to identify past trends, predict a traffic spike, spin up new software-defined appliances, and route traffic accordingly. The result is no service degradation or disruption and an optimal experience for the end users.
Some modern provider networking platforms provision hardware that costs per hour of usage. Using AI to determine what the next eight hours of network traffic will likely entail enables manual or automated action that can release that hardware and end your financial obligation. Alternatively, if an organization owns the hardware, it can still save money and encourage innovation by freeing up compute resources that can be used in some other meaningful way instead of sitting dormant.
Meraki AI-Driven Traffic Management
The Cisco Meraki platform, the leader in cloud-controlled Wi-Fi, routing, and security, relies on cloud-based management and enhanced telemetry. This design enables it to collect and process large amounts of information from connected devices at numerous locations. The data is transmitted to Meraki cloud servers where machine learning models perform real-time analysis. These algorithms detect deviations and patterns that inform present network status and predict future trends.
Meraki’s AI-powered Auto RF uses AI-enhanced Radio Resource Management (RRM) to optimize radio parameters for high-performance and reliable wireless connectivity. This feature enhances the experience by personalizing the network and understanding its trends, especially in an enterprise environment.
Key features of Auto RF include:
AI Channel Planning, which avoids channels that are more susceptible to interference.
Busy Hour, which prevents unnecessary adjustments during peak times.
With these innovations, Meraki can offer seamless and efficient wireless networking that is suited and adapted to these networks’ distinctive requirements and circumstances. Meraki also features a root cause analysis feature powered by AI that enables advanced troubleshooting and debugging of network issues. The Meraki dashboard continuously collects and analyzes telemetry data from various network devices, using AI algorithms to detect anomalies and performance degradation. When an issue is detected, the Machine Reasoning Engine performs root cause analysis to identify the underlying cause of the problem.
As shown in the figure, the Meraki dashboard has identified an ISP issue at a local branch. The AI-powered root cause analysis finds that VPN backhaul is the cause with medium confidence. The produced “evidence” reveals that seven clients are affected by traffic being sent over VPNs. This information enables network managers to precisely locate where network issues are occurring and gives them insights into possible solutions they could act on.
Last updated