Selective and on-demand network measurement with SRv6 and INT

Network measurement plays a critical role in network management to provide rich real-time network state for further network analysis. The novel network measurement approach, In-band Network Telemetry (INT), provides real-time and fine-grained packet-level network measurement. However, INT incurs significant overhead and lacks of flexibility to perform configurable on-demand network measurement. In this work, by combining the programmability of SRv6 and the telemetry efficiency of INT, we propose a Selective On-demand Network Measurement mechanism (SONM-SR-INT) to enable customizable network measurement where the measurement targets and telemetry information can be both configured. SRv6 is applied to both drive the measurement probing packet to follow a certain probing path to cover a given list of network targets, and facilitate the upper layer INT to perform per-hop customized network telemetry. We determine the optimum probing path by solving a Traveling Salesman Problem (TSP) in an auxiliary graph. We also define the SONM-SR-INT probing packet frame and packet processing mechanism. Finally, we implement SONM-SR-INT in P4 and validate its performance through a set of experiments.