Cloud-Native Networking and eBPF/XDP-based Packet Processing

This research aims at realizing a service infrastructure that supports next-generation networks. Cloud-native network functions (CNFs) based on containerization accelerates network service deployment with agility and flexibility. For example, a network service is realized by deploying the network functions required by the service requirements on distributed systems. Because the network functions has a complex nature, it is difficult to monitor network metrics more efficiently and accurately than in conventional monolithic systems. Therefore, we aim to efficiently collect highly accurate network metrics from the distributed systems using a program running on the kernel called eBPF. The data obtained is used to predict a failure in the distributed system and detect abnormal traffic. Furthermore, intelligent and fast packet processing can be achieved by applying artificial intelligence and machine learning to eBPF/XDP-based kernel networking.