Computer network outages and quality of service issues can be costly and have negative effects on businesses, including lost revenue, lost productivity, poor customer and employee experiences.
In traditional networks, when logical or physical failures occur, the network is most likely programmed to automatically convert to a standby state.
However, this and similar forms of redundancy, along with the methods that support them, are often seen as less effective, as they offer only limited insights into key issues − the most important of which relate to affected traffic flows.
Today, as technology advances and increasingly complex corporate network and infrastructure goals are achieved, network users and administrators are realising the contributions of artificial intelligence (AI), machine learning (ML), data analytics and automation technologies as they apply to redundancy and uptime maintenance.
In essence, network administrators are appreciating the relevance of these technologies in their efforts to maintain fully-functional and productive corporate networks.
The roles played by AI, ML and associated technologies are giving organisations the ability to boost – and even ensure − uninterrupted connectivity and business continuity in the event of network outages.
Self-healing networks hold the promise of a future-proofed infrastructure underpinned by reliable performance.
This heralds a paradigm leap in network resilience and signals the arrival of self-healing networks.
For organisations, self-healing networks hold the promise of a future-proofed infrastructure underpinned by reliable performance no matter how today’s interconnected world evolves.
What is a self-healing network?
According to prominent US-based author Dennis Burger: “Self-healing networks sound a bit magical, but are like most, if not all, of the technological innovations permeating our lives these days − which is to say that they’re certainly a convenience, but not without some consequences.”
Most definitions focus on the technology linked to the self-healing network and its built-in properties to protect against failures by predicting incidents, identifying problems or faults, providing solutions or work-arounds, and supporting automatic recovery without human intervention.
Andrew Froehlich, a US-based IT industry executive, says the value of a self-healing network depends on how effectively technologies such as AI, ML and others are applied.
“The goal of a self-healing network is its ability to identify, remediate and predict problems. In so doing, the operation of the network must be simplified to the point where it frees those operating it to focus on more high-value tasks,” he notes.
Realistic expectations
Despite the broad-based acceptance of AI and the enthusiasm with which the idea of self-healing networks has been welcomed, there is one vital question still to be answered: Can self-healing networks be trusted?
Trust is earned in many ways and depends on a number of factors, including the accuracy of the underlying AI models, the robustness of security measures put in place and the levels of transparency provided.
Trust is multi-layered and is often defined as a spectrum that is conditional on how well self-healing networks are planned, designed, commissioned, developed and maintained.
Network safety is probably the most important aspect of the trust spectrum. Networks with self-healing capabilities need to be protected from hostile attacks which can be launched from anywhere and arrive in many forms.
For example, AI's decision-making processes have been known to be vulnerable to manipulation by bad actors, leading to destructive outcomes.
In this light, AI models need to be robust enough to handle security challenges that may be compounded by the growing complexity and diversity of modern networks, which include an ever-widening range of devices, protocols and applications.
An effective communications strategy is also a key trust factor. Should AI models be unable to communicate or co-operate efficiently with current systems, the creation of a fully-functional self-healing network will most likely be jeopardised.
Although AI-powered self-healing networks frequently function with little to no human supervision, the possibility of human intervention – as might be required in certain circumstances − can be crucial in engendering trust.
More specifically, trust can be fostered through a “human-in-the-loop” mindset facilitating human-AI collaboration. The chances of poor decision-making by AI systems can be reduced, if not eliminated, through the incorporation of human override options and fail-safes.
In a similar vein, ongoing performance monitoring guarantees the systems are operating efficiently and are able to adjust to new situations and challenges as they arise. Feedback loops also help AI systems evolve by allowing them to learn from past errors and omissions.
Trust can also be maintained in self-healing networks through accurate and thorough record-keeping. This will allow administrators to fully comprehend the rationale behind AI decision-making and understand how these systems arrive at their choices. As a result, administrators should be able to put more faith in the outputs of AI systems.
Finally, any ethical questions surrounding AI need to be addressed and answered. Ethical principles and societal norms need to be considered and all instances of bias in AI models must be identified and eliminated. Impartiality and equitable decision-making must be guaranteed.
How well these aspects are addressed will influence the level of trust generated by AI-powered self-healing networks and determine their overall effectiveness.
Looking to the future, it is reasonable to believe network users and administrators will come to implicitly trust AI to make important decisions. This will give self-healing networks the momentum to develop into extremely dependable and trustworthy elements of contemporary IT infrastructure.
It may take a little time but, as noted by Froehlich: “Once that trust is earned, the true power of self-healing networks can finally be realised.”
Share