Marvis and the future of smart network management

Organisations are increasingly confronted with complex network issues that can disrupt operations and hinder productivity. Traditional troubleshooting methods often fall short. This leads to prolonged downtimes and frustrated users. Enter artificial intelligence (AI), a transformative technology ready to change the game in network management.

By leveraging AI, businesses can proactively identify and resolve network problems, ensuring seamless connectivity and enhanced user experiences. Changing from reacting to problems to preventing them helps solve issues faster and boosts network performance. This makes AI an important partner in tackling today's networking challenges. One such AI-driven solution is Marvis.

Who is Marvis?

Marvis is a cloud-based artificial intelligence (AI) virtual network assistant (VNA) developed by Juniper Networks that is designed to enhance network management and operations. This innovative tool reflects a significant advancement in how networks are designed, deployed, and maintained, drawing on years of research and development in networking.

The evolution of network management

The inception of Mist Networks marked a pivotal moment in network management. Founders Sujai Hajela and Bob Friday recognised a fundamental shift was necessary to meet the demands of modern networking. Traditional management tools were struggling to keep pace with the growing complexity of network environments. A solution for smooth user experiences from client to cloud is needed, no matter what the network components are.

This realisation led to the creation of Mist, a modern network platform that uses cloud technology and AI to improve user experiences.

Introducing Marvis and cloud AIOps

Marvis emerged from the vision of creating an AI solution capable of addressing networking challenges with the same proficiency as human experts. The team was inspired by IBM's Watson, which won on Jeopardy (an American television game show). They wanted to create a system that can analyse network data, understand language queries, and automate tasks.

Marvis operates on the Juniper Mist AI microservice cloud platform, processing vast amounts of real-time data from various network elements. By utilising deep learning and natural language processing (NLP), Marvis provides actionable insights and automation, significantly improving network operations.

Key benefits of Marvis

The implementation of Marvis has led to several notable advantages for its users:

• Reduced support tickets: With handling many common inquiries and issues, the volume of support tickets has decreased.
• Faster troubleshooting: By employing Service Level Expectations (SLEs) and advanced deep learning models, it can quickly identify and resolve issues.
• Enhanced user experience: The virtual network assistant continuously measures and optimises user experiences across various applications, including voice and video services.
• Operational efficiency: IT teams benefit from increased efficiency, allowing them to focus on strategic initiatives rather than routine troubleshooting.

It is not just a product; it is a platform that grows and adapts by learning from new data and the network's needs.

Training Marvis for success

The effectiveness of Marvis is rooted in its training methodology. A close collaboration between the data science and customer support teams has been essential. Regular meetings allow for the identification of areas where the virtual network assistant can improve, effectively employing reinforcement learning through human feedback. Reinforcement learning is a type of machine learning in which an agent learns to make decisions by interacting with an environment.

The training process involves sophisticated techniques, including supervised and unsupervised learning, enabling Marvis to diagnose a wide range of network issues. Additionally, natural language understanding (NLU) capabilities empower Marvis to interpret user queries and respond appropriately.

Marvis uses reinforcement learning to solve network problems independently. It can also seek human assistance when necessary to ensure effective responses.