Mathematical Models for In-Band Telemetry Orchestration

In-band telemetry establishes that signals for computer network monitoring can be sent via user data packets. As the space in these packets is limited and the monitoring applications are heterogeneous in relation to the monitored characteristic, requiring different information for the correct operation, the combinatorial problem is to decide which set of telemetry items optimizes the overall management of the network. The In-Band Telemetry Orchestration Problem has variants focused on reducing the overhead caused by including telemetry data in streams. In this work, two variants of the problem are studied, focused on monitoring applications and their respective mathematical models. The proposed models allow you to optimize instances with 50 and 100 network devices and 50, 100, and 150 flows. Comparative computational results are presented with a model that considers the maximization of telemetry collection as an optimization criterion.