Network-Intelligence (NI) Project
Network Intelligence is a research project supported by Korean government* (MSIT/IITP) to develop Network Function Virtualization (NFV) lifecycle management functions (see below) including physical and virtual resource management in NFV environment using Artificial Intelligence (AI) technology. The main goal of this project is to develop an NFV management and orchestration (MANO) platform that can support network operator’s management decisions and eventually operate the NFV environment in an autonomous way, using AI and Machine Learning (ML) with massive data from underlying infrastructures. This also covers automation of provisioning and verification process of such decision making by AI/ML. This research project has been started since July, 2018.
* [2018-0-00749, Development of virtual network management technology based on artificial intelligence]
Research Topics in Network-Intelligence Project
1. NFV Monitoring
- Development of efficient monitoring technology for virtual networks and physical / virtual resources constituting NFV.
- Since NFV is basically deployed in cloud computing environment, it not only monitors status of servers (including hypervisor) and networks but also traffic information among Virtual Machines (VMs) and VNFs running on top of them. The NFV monitoring function collects these types of information in real-time and provides the ability to convert the collected data into a form suitable for machine learning.
2. VNF Deployment
- Ability to deploy VNFs (application functions) with various requirements (user, provider, etc.) in NFV environment that is composed of different types of virtual servers and virtual networks. This function can use machine learning (also deep learning) technology, in order to place VNFs with satisfaction of various multi-objectives such as ensuring lowest latency or highest utilization by considering characteristics of different VNFs.
3. Service Function Chaining (SFC)
- Ability to chain multiple VNFs according to the service requirements of the network user. VNF is developed by using machine learning technology to minimize the load and link bandwidth and delay of virtual server nodes and to dynamically respond to faults.
4. VNF Resource Demand Prediction
- The ability to anticipate the demand for resources that the VNF will require. Developed with machine learning technology using VNF and other VNF information connected to it.
5. VNF Auto-scaling
- VNF The ability to dynamically increase or decrease the resources (virtual servers) allocated to VNFs based on resource usage and traffic load. We developed an auto-scaling function with machine learning technology that guarantees user’s QoS / QoE without sacrificing performance during scaling and optimizes NFV resource usage.
6. Anomaly Detection and Attack & Intrusion Detection
- To prevent degradation of performance of NFV, it detects and notifies abnormal behavior of VNF or virtual and physical server in advance, and detects and protects abnormal symptoms such as attack or intrusion of NFV platform using machine learning technology box.
7. VNF Live Migration
- VNF Live Migration function that predicts the possibility of overall failure and moves VNF deployed virtual machine in real time. This function is developed based on machine learning technology, minimizing the down time of VNF and resolving IP address change problem that can occur when moving virtual machines in WAN environment based on LISP protocol.
8. NFV Power Management
- Based on machine learning technology, we developed electricity power management function that reduces power cost by driving VNF in minimum number of physical nodes within the range that meets service level agreement (SLA) of all VNF running in NFV environment.
9. AI-based NFV Management and Orchestration Platform
- Development of an NFV-MANO platform that integrates AI-based NFV lifecycle management functions above to perform functional and performance testing and verification of Machine Learning models.