DINT: A Dynamic Algorithm for In-band Network Telemetry

Abstract

Network monitoring is fundamental for the correct and expected functioning of today’s large computer networks. In-band Network Telemetry (INT) has become one of the main tools for collecting network information in recent years. By piggybacking information using business packets, INT can deliver real-time network statistics to monitoring engines. However, INT’s fine granularity comes with a high network overhead cost. This paper focuses on balancing this trade-off between accurate monitoring and high telemetry overhead. To achieve it, we propose DINT, a Dynamic INT algorithm capable of adapting to different traffic patterns while keeping an accurate view of the network and reducing flooding it with redundant telemetry data. In our experiments, DINT presented higher adaptability compared to other techniques, providing a more accurate view of the network while requiring fewer telemetry data.