Nikhil Handigol is Co-Founder and Chief AI Officer, Forward Networks, building the foundation for autonomous networking.
getty
As networks have grown larger and more complex, one aspect has not evolved. We still operate the backbone of our businesses without a provable understanding of how it actually behaves, and we implement production changes without evidence of their impact.
In any other domain, this would not be tolerated. The pharmaceutical industry models molecular interactions before clinical trials. Engineering runs unit and integration tests in a staging environment before deploying systems. Yet in networking, where operational resilience is directly tied to business success, changes are still implemented based on outcomes that cannot be verified before they happen. And it’s (begrudgingly) accepted.
Outages and security incidents are most frequently caused by erroneous changes made during maintenance windows. This is true across industries and company sizes, and the financial consequences are significant. Gartner estimates (registration required) that a major unplanned network outage costs more than $500,000 per hour, with total business losses that can exceed $1 million.
Data is not the problem. Network engineers are inundated with data from tools built for monitoring, change management, configuration management, incident response and more. These tools were designed to help networks scale and improve resilience. In practice, they produce fragmented views and constant alerts. What they do not provide is a complete and provable understanding of how the network behaves across the entire estate. Without that, there is no reliable way to ensure that a proposed change will deliver the intended behavior in production.
This is the operational gap that networking teams have been working around for decades. Change management processes aim to align production behavior with intent, but without a provable understanding of how the network actually behaves, every change carries risk that should not exist.
Other teams do not operate this way. Software is comprehensively tested in a staging environment before it’s delivered, and cloud teams use dev test environments to prove outcomes before changing the production environment.
Networking hasn't kept pace with that standard. The change process still relies on developing a method of procedure, testing in a limited lab, reviewing through a change review board (CRB) and executing in live production based on an incomplete understanding of network behavior and tribal knowledge. The gap between intent and the production reality remains. Outcomes cannot be proven in advance, and the result is that critical infrastructure is changed based on informed guesswork.
At the same time, pressure on IT leaders to adopt AI and autonomous networking is accelerating. AI agents are being deployed to manage routine tasks at machine speed. But speed applied to an unverified environment does not create efficiency. It accelerates the rate at which unintended outcomes reach production. The operational gap does not disappear with AI. It scales with it.
Change processes today rely heavily on manual verification and introduce security, compliance and operational risk. The gap between intent and production must be closed. Doing so requires a fundamentally different standard of understanding, grounded in a mathematically accurate understanding of network behavior.
A mathematical model of the network enables a definitive analysis of how the network behaves. This includes proving network isolation, verifying behavior, validating policy compliance and creating the guardrails to safely enable autonomous networking.
In a mathematically accurate digital twin, every network device is modeled as a transformation function on the set of all possible packets it can process. When these transformations are analyzed end-to-end, the full network can be verified against the intended behavior and policy.
This level of precision forms the basis of what is often referred to as a network digital twin, built from the actual configuration and state of every device across the hybrid network. It enables outcomes to be understood and predicted deterministically with a level of accuracy that lab testing cannot deliver.
Autonomous networking requires that every action be verified against a mathematically accurate model of the full production network before execution. Without that, the unintended outcomes, such as broken connectivity, security exposure or unintended access, cannot be prevented.
Eliminating the operational gap redefines what is possible. Autonomous networking can be implemented with confidence because every action is verified before it executes, rather than after the fact. Changes succeed on the first attempt. Change cycles that previously stretched across days or weeks are successfully completed in hours.
Highly skilled engineers are no longer spending hours on manual verification. Instead, they can focus on strategic architecture, optimization and infrastructure projects that drive the business forward.
New services, cloud expansions and AI workloads are no longer delayed while the network catches up. The network enables the business to move faster, supports autonomous operations and operates with a level of certainty that was not previously possible.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。