Title: Self-driving Networks
Technical Area: Networking
Large scale networks are challenging to manage. The life cycle of network management, including network planning, configurations, monitoring, diagnosis, debugging, security, updates, data analytics, etc., is essential to the reliability, performance and cost for operating a network. Traditionally, the whole network management life cycle is intercepted and controlled by human in a manual or semi-manual fashion. For instance, a new network is usually designed by human, and troubleshooting is mostly based on human intuitions and decisions. This situation is not sustainable given that human decisions and operations can be short-sighted and error-prone in complicated networks, while, to make it worse, networks in large online service providers like Alibaba are rapidly growing in size, speed and application demands. Therefore, it is critical to introduce machine intelligence into network management.
The ultimate goal of this project is to make networks operate itself without human involvements 24/7 with high reliability, security and efficiency. For achieving this goal, we call for collaborations from researchers in academia to introduce artificial intelligence (AI) into the network management life cycle, either for helping human to make better decisions or even “drive” the network automatically.
Related Research Topics
The suggested topics include but not limited as following:
- Network planning with AI
- Applications of machine learning to network incident prediction and remediation
- Practical network verifications on both control and data planes
- Network measurement techniques that adapt collection or measurement
- based on changing network conditions
- Query languages that support queries for network status
- Predictive machine learning approaches to closed-loop network management systems
- Design and implementation of closed-loop feedback controls for the combined detection and mitigation network bugs
- Closed-loop network management systems that can incorporate human feedback to achieve better reliability, performance, security and cost efficiency