Artificial Intelligence (AI) is all the rage. It is widely promoted as the panacea for improving business performance across a wide range of different areas. How, specifically, does AI enhance the management of computer networks?
The term AI has its origins in Science Fiction, often linked to far-fetched tales about machines attempting to take over the world. Now that AI has become a reality, the concept has changed, but the basic premise of computers that are capable of making independent decisions lies at its heart.
In essence, in the context of IT systems, Artificial Intelligence analyses vast amounts of data and provides insight based on its findings. When coupled with machine learning, it can be used to automate decision making in response to different events and perform tasks with no need for human intervention.
The appeal of AI in network management
In computer networks AI and machine learning are now being used to continuously analyse large quantities of data using sophisticated algorithms to determine what exactly is happening on the network, to make predictions and to respond to events as they happen. This ability to intelligently analyse network data and gain detailed insight into the network’s performance with no human intervention lies at the core of its appeal.
The essence of AI, and the reason it is gaining so much attention across the entire IT world, is that it enables intelligent automation of many tasks, saving vast amounts of time while also improving operational effectiveness. This applies to network management where many of functions involved in the efficient operation of a network can be automated, dramatically improving network performance, troubleshooting and security.
One simple example, a network switch function that has been around for a long time, is loop-back detection. It’s a feature on smart-managed and managed switches that has saved network administrators tremendous amounts of time in the event of accidental or intentional network misconfiguration. A network loop occurs when a cable from one switch port connects back to another port in the same switch or network. Loop-back causes a broadcast storm that brings the network to its knees because network traffic is continuously amplified rather than stopping at its intended destination. With loop-back detection, when this occurs, one of the affected ports is automatically shut down, mitigating the problem. Without loop-back detection, the network administrator has to manually locate and correct the fault that could be anywhere across the entire network.
AI and machine learning result in reduced downtime, pre-emptive maintenance and reduced operational costs while, at the same time, saving network administrators’ time. AI’s evolving role in this space is making it easier for companies – particularly SMBs – to more effectively run their networks and is bringing us closer to self-healing networks and zero-touch network management.
Utilising network log data
In any computer network, there are vast amounts of machine data continually generated by internal processes and through server logs, Wi-Fi controllers, applications, connected devices and other networking equipment. In a conventional network setup, much of this data accumulates in logs and is rarely accessed. The introduction of AI and machine learning enables network management systems, through automation, to interpret this data, determine what is happening in fine detail and use this insight to continuously improve network performance and reduce downtime. Furthermore, it does this more quickly and more accurately than humans ever could.
AI and machine learning can be used to detect problems and apply solutions to common network issues without human input making it a powerful tool in maintaining and improving network operations. In a Wi-Fi network, for example, this may mean maintaining full network coverage in the event of the failure of an Access Point (AP) by automatically increasing the RF signal strength of other Access Points (APs) in order to reconfigure the network and cover any potential dead spots.
Currently, the focus is on AI and machine learning taking on the more administrative and mundane areas of network management. Essentially, teaching the network to automate basic management tasks and alert network administrators if more complex problems are identified that require human intervention.
Automatically prioritising critical network traffic
AI built into smart switches is now also being used to guarantee the timely delivery of critical traffic as it flows across the network. By analysing Ethernet packets, these intelligent switches can automatically assign different levels of service to different types of network traffic and prioritise IP video and VoIP packets without compromising the transmission of other network data. This saves the expense of having separate, dedicated hardware specifically for IP voice and video.
Using a technique known as Auto Surveillance VLAN (ASV), real-time IP video packets are given priority so that the quality of real-time video for monitoring and control is assured. Similarly, Auto Voice VLAN technology guarantees the quality and security of VoIP traffic and ensures uninterrupted VoIP calls for network users.
AI and cloud-managed networks
Network architectures are increasingly moving to a centralised management structure with management functions handled in a control plane that is separate from the data plane, such as with cloud-managed networks and Software Deﬁned Networking (SDN). AI and machine learning are essential to reaping the full benefits of these centrally managed network architectures.
Automated analytics and learning capabilities are needed to realise the benefits of improved network flexibility and ease of management that such infrastructures deliver. The combination of AI and centralised software-based network management is driving us towards fully automated networking.
The most significant area where automation is currently being used is in zero-touch network configurations with intelligent network devices that automatically connect to a server on power-up for automatic configuration and updates. This AI-driven ‘plug and play’ deployment removes the need for extensive manual setup, saving time and enabling devices to be deployed at remote locations without on-site network administrators.
AI tools are also increasingly being used to improve network monitoring, management and analytics, driven by their value in predicting network issues and automating fixes before they become problems. The ability to interrogate massive amounts of data from multiple sources means that AI and machine learning provide better insight into day-to-day network performance and current utilisation levels. This increased insight into what is happening on the network means that changes and challenges can be detected early and proactive action can be taken to ensure performance is continually optimised. Identifying traffic patterns and understanding network trends means that more accurate forecasts can be made, significantly improving the accuracy of network capacity planning.
AI and machine learning are also playing an increasingly important role in improving network security, giving better insight into behaviour on the network so that threats can be automatically identified and acted upon quickly.
Of course, as an automated approach, AI and machine learning work round the clock, providing continuous and constant management and control of the network with no down-time meaning additional service improvements.
As networks continue to grow in complexity, adopt centralised management architectures and support a broader range of connected devices and operating systems, AI-powered network management is central to streamlining, troubleshooting, and improving the operation of the network. We are fast approaching the time when AI and machine learning are essential to the smooth running of any network.