Network Copilot™ Use Cases

There are many possible use cases for Network Copilot™, but the main ones include:

① Network Update
It is possible to suggest the appropriate timing for updates based on traffic increase trends and easily generate performance reports after updates.

② Security Audit
You can load your company's security requirements in advance, inquire via chat whether those requirements are met, and even integrate it into your company's automation tools.

③ Performance check
You can request the creation of the dashboard you want via chat, and grasp the status of your network devices in real time.

④ Operational standards check
You can also keep up to date with the latest operational standards, chat with us to ensure you're aligned with them, and integrate Network Copilot™ capabilities with your existing NetOps tools.

⑤ Troubleshooting
Even beginners can contact the system via chat and receive real data retrieved from the network and suggestions for resolving similar problems that have occurred in the past.

Network Copilot™ System Architecture

Next, we will introduce the system architecture of Network Copilot ™.

The green blocks utilize the latest open LLMs and libraries. The current LLM uses a 7B (7 billion) parameter model from Mistral, a French company.

The blue blocks are parts that Aviz is developing for Network Copilot ™.

We collect data from your network and store it in our database. We also collect your operational standards and knowledge base, and can provide answers based on the information referenced in the RAG method.

When a user makes an inquiry in Japanese regarding network operations, Network Copilot ™'s LLM queries the database and returns a response based on the results.

To better understand the operational flow of Network Copilot ™, let us take a look at the following example to see how it works.

As an example, consider the case of enquiring whether a network meets operational standards.

The operational standard is that a network device's CPU load is considered normal if it is below 80%.

First, as a preliminary step, constantly import CPU load data from network devices into a database. Also import a file that lists your company's operational standard of a CPU load of 80% or less.

In a real world query, a user asks, "Is my network meeting operational standards?" The LLM understands that query and generates a SQL query to query the database, which in turn retrieves CPU load data from the database. Let's assume that the load is 50 %.

At the same time, the RAG method is used to reference information that states that a CPU load of 80% or less is normal according to the company's operational standards, and the system asks the LLM, "Currently, the CPU load is 50 %, and the operational standard is that a load of 80% or less is normal. Does this meet the operational standard?" The LLM then replies to the user, "Based on the acquired data and the operational standard, the standard is met."