Even though I feel like I’m well versed in networking there are times when I feel like I’m clueless. It happens when someone asks me how to find important information in a system that I’m not familiar with. I may have spent a large part of my career working with Cisco routers or Aruba access points but if you ask me to find information on a Juniper switch or a Palo Alto firewall I’m going to be struggling to get around. I may know what I want to find or what I need to accomplish but translating that request into something that is able to be executed on the unfamiliar device is a challenge in and of itself.
One of the promises of AI is that it can do the translation for you. I’ve seen countless examples of LLMs and GPT algorithms providing essays for homework and code that looks like it should pass muster when pasted into an IDE. However, I don’t fully trust the output because I didn’t see the steps to get there. I typed a prompt and something came out the other side. Did the system interpret my prompt correctly? Is there some nuance to the language that it missed? Perhaps it was something that I phrased incorrectly that led to the wrong answer. No matter what it was I feel like obfuscating those intermediate steps is a hazard we need to avoid at all costs.
Fast Forward To the Fun Parts
Recently, Forward Networks released a new AI Assist feature for their platform that has me excited. I got a chance to see it in action during a demo at Networking Field Day 34 when Nikhil Handigol showed how it works. If you’re not totally familiar with Forward Networks there’s a lot of other great content in the presentation. The short version is that Forward builds a digital twin of your network that allows you to query the model for details about devices and understand interactions so you have a complete picture of what your systems look like.
The star of Forward’s platform is the Network Query Engine (NQE). This is a powerful way to ask questions to get information from your network. NQE allows you to ask about devices and code levels and paths and other critical information. All the things you want to know right at your fingertips. Well, almost at your fingertips. Because NQE is still a language. It has a syntax and a way to structure your query to make sense to the underlying system. Like the above example, you need to know how to format your request into the method the system uses to extract data. If you know you want to see the code levels on all the routers in a given site you need to know what to ask and how to narrow the query so it gives you exactly what you’re looking for.
Anyone with a background in databases knows how hard the process can be. Anyone that has every tried to learn SQL will tell you it’s easy to get the basics but as you add layers of jointing associate entities and filtering information you add layers that not only create failure points but also impact the way the data is presented. If you don’t craft the query in just the right way you could find yourself missing critical data points or incorporating information that is absolutely useless and just muddies the waters.
That’s where the new AI Assist functionality comes into play. You can ask the Forward Networks platform a plain language query and it will go into an LLM and out comes the query you need to run to find the data. Note that it doesn’t produce the results of the query. Instead, it takes your prompt and outputs the NQE syntax you need to use to find the information. Yes, it adds an extra step compared to something like ChatGPT but it’s a very crucial step for a professional setting.
Because AI Assist shows you what you’re about to type in it gives you a chance for a sanity check. You may not be totally familiar with the NQE syntax but being able to see what you’re looking for helps you understand the various fields that can be searched as well as how the query is put together. Even with rudimentary knowledge of the syntax you should be able to spot if something is out of place. More importantly, seeing the query before you run it means that you’ll be able to learn the language faster and understand why it produced a given output.
Bringing It All Together
This is a wonderful way to implement an AI function in a platform that provides information about a network or a collection of systems. Rather than just hoping the software is giving you the right data you can audit the query and make sure it’s returning the right results. You get to ensure that you’re getting high-quality information from your system and even be able to tweak the parameters to gain more insights. It’s the way that all systems should help you get to the data you need. Holding your hand at the start but opening up the full capabilities to those that want to put in the time to learn the power under the hood.
For more information about Forward Networks, NQE, and their new AI Assist feature make sure to check out their website.
