Blackbird: A Scalable and Resource-Efficient Framework for Distributed Network Monitoring

Elisa Arnoldi, Gabriele Merli, Marco Abbadini, Michele Beretta, Dario Facchinetti, Matthew Rossi, Stefano Paraboschi

Modern large-scale network infrastructures require monitoring solutions that ensure operational resilience while imposing minimal overhead. However, existing centralized and static multi-agent approaches often suffer from single points of failure and can generate significant traffic congestion.

This paper introduces Blackbird, a distributed framework designed to provide a continuous, real-time view of the entire network. Blackbird transforms selected network hosts into self-organizing agents capable of autonomously performing network scans. Within this architecture, a subset of agents act as aggregators, responsible for scheduling scans to mitigate burstiness and coordinating the activity of workers. Workers, in turn, perform measurements and disseminate the collected data across the system. Preliminary experimental evaluation shows that Blackbird achieves robust horizontal scalability and strong fault tolerance under node failures. Moreover, it maintains a low CPU and memory footprint, which is especially important for resource-constrained edge devices, while preventing network saturation and enabling real-time inspection from any participating node.

Paper

@inproceedings{blackbird,
  author = {Elisa Arnoldi and Gabriele Merli and
            Marco Abbadini and Michele Beretta and
            Dario Facchinetti and Matthew Rossi and
            Stefano Paraboschi},
  booktitle = {Proceedings of the 46th IEEE International
               Conference on Distributed Computing Systems
               (IEEE ICDCS 2026)},
  title = {Blackbird: A Scalable and Resource-Efficient
           Framework for Distributed Network Monitoring},
  year = {2026},
}