Bubbln: The Background Story

Download Bubbln v1.0.0

Over the years, setting up networks required engineers logging into each networking device such as routers and switches through remote accessibility tools like Secured Shell (SSH), and then use Command Line Interfaces (CLI) to manually key in series of commands. This method is prone to errors, especially when setting up large scale networks which can be monotonous and time consuming. As networks grew in complexity, the need for quicker and more efficient setup processes, along with the desire to overcome the challenges associated with manual setup, gave rise to the concept of network automation.

Network automation is the use of software to automate tasks normally performed by network engineers. These include configuration, provisioning, monitoring etc. This approach offers several advantages over traditional methods, including improved efficiency, reduced errors, increased agility, and enhanced security. However, despite these benefits, acquiring the skills needed to operate and manage the automation tools can be time-consuming.

Recently, there has been advancements in Artificial Intelligence, particularly Large Language Models (LLM) e.g. ChatGPT, Bard etc. which has led to a quick and wide range adoption of LLMs, as well as application in various areas such as software coding, data analytics, web design etc. This is largely due to its ability to process texts in natural languages to generate desired text outputs.

Considering this capacity of LLMs which can eliminate the technical skills requirement of current automation tools, Bubbln was developed (as a Master’s degree research work) to explore the potential of using ChatGPT as an API-based network automation tool. The following research questions would serve as guidelines to meeting the project’s objective.

• Is ChatGPT capable of generating Ansible playbooks?
• What is the accuracy of the generated playbook?
• What is the accuracy of the generated playbook?
• How efficient is ChatGPT to automate a network compared to Ansible framework?
• What is the consistency and reliability of generating playbooks across different scenarios?
• How easy is it to use ChatGPT to automate a network?
• How much computing resource is utilized to enable ChatGPT automate a network?
• How scalable is the use of ChatGPT to automate a network?