Can AI Operate the Internet?

Skynet? No, no Terminator, please.

Just joking—don’t jump to the movie scene right away. Let’s consider the idea of AI operating a network, similar to how automation handles network device configuration and zero-touch provisioning. Is it possible?

Let’s start with automation. A device comes with an initial configuration that includes an IP address and gateway. Once it’s network-accessible, the central server recognizes the device and pushes the final configuration. That’s a simple flow, right?

But what happens next—who operates the device?
Today, it’s still a human network operator.

Could AI replace the network operator? It depends on how the AI model is trained, how it handles fault tolerance, and how it avoids operational errors. There’s no such thing as 100% guarantee.

However, using AI to operate parts of the network—such as traffic tuning, AS path changes, or metric modifications—seems realistic.

AI can assist with tasks like network measurements, route and traffic tuning, DDoS detection and auto-mitigation (which already exists), device fault detection and traffic redirection, and DNS updates. That all seems feasible. You could train an AI module specifically for this purpose.

But then again—do we need AI if existing technologies already meet these requirements?

Let’s consider a scenario and see what AI (specifically, ChatGPT) can do.


Scenario:

  1. An Internet provider has 2 transit links:
    • Link A: 1 Gbps
    • Link B: 10 Gbps
  2. The subnets 10.0.0.0/24 and 11.0.0.0/24 (AS 65000) are advertised to Link A.
  3. The same subnets, but with AS path prepending (65000, 65000), plus 12.0.0.0/24 (65000), are advertised to Link B.
  4. Link A is congested—11.0.0.0/24 is using nearly 900 Mbps.
  5. Link B’s total usage is just 5 Gbps.
  6. A customer on 10.0.0.0/24 complains they cannot access the Internet.

ChatGPT’s Suggested Operation:

Summary:

  • Remove or reduce AS path prepend on Link B.
  • Optionally add prepends on Link A.
  • Let traffic shift to Link B (10 Gbps).
  • This will relieve congestion on Link A.
  • It should also restore Internet access for users on 10.0.0.0/24.

Business Follow-Up:

“From a technical perspective, that works. But from a business point of view, Link A is now underutilized. If Link B has a higher cost per bps, this is inefficient.”

ChatGPT’s Response:

AspectStrategy
Technical QoSOffload heavy prefixes like 11.0.0.0/24 to Link B
Cost ControlKeep lighter prefixes like 10.0.0.0/24 on Link A to avoid overusing B
Business FitAvoid paying premium for B unless capacity is needed
Scalable OptionAutomate prepend adjustments using monitoring-based thresholds

Yes, that’s a great example of how AI can analyze your needs if you provide enough detail and criteria.

The remaining concern is AI’s operational cost and power consumption.

This topic definitely deserves more discussion—SD-WAN + AI?

#BGP #bgptrace #AI #chatGPT #AS #ASPrepend #path #link #transit #INTERNET #SDWAN

Looking glass function provided by RIPE Atlas?

I performed some traceroute tests using the public looking glass of another organization/provider. I found that some test functions, like Ping and Traceroute, were launched using RIPE Atlas probes. It looks impressive and kind of funny.

In the previous year, the provider developed a web interface and API to launch commands from their own PE (Provider Edge) or Internet BG (Border Gateway) routers and return the results. The geographical router list allows users to select region-based tests.

This seems to be a new method using RIPE Atlas, where queries can be made via an API. The web interface lets users select which probe to use for the measurement, deducting the web provider’s “RIPE Atlas Credits” for each test.

However, I’m wondering — since looking glass aims to provide insights into a specific network provider’s or AS owner’s network — if we’re using this method, why not just go to the official RIPE Atlas website to launch the test?

Well, I guess the more user-friendly web portal makes it easier for users.

Pingnetbox – http://www.pingnetbox.com

#ripe #atlas #lookingglass #measurement #ping #traceroute #test #internet #AS #chatgpt #proofreading