QoS Diagnosis in Networks Based on Joint Analysis of Survival and Statistical Changes in Performance Time Series

• Ian José Agra Gomes UFRJ / CASNAV
• Gabriel Buginga UFRJ
• Edmundo de Souza e Silva UFRJ
• Rosa Maria Meri Leão UFRJ
• Gaspare Bruno Anlix

DOI: https://doi.org/10.5753/wperformance.2025.8604

Abstract

This work presents a novel methodology for assessing the quality of service in computer networks. It combines time-series change-point detection techniques with clustering based on survival analysis. Motivated by a partnership between M-Lab and RNP, the methodology was applied to real data collected from an ISP using the NDT protocol. It allowed the identification of clients and servers with poorer performance than others. To facilitate the interpretation of the results by non-specialized teams, we used a large language model, which produces understandable reports from the methodology results.

References

Abdin, M. I. and et al. (2024). Phi-4 technical report. Technical Report MSR-TR-202457, Microsoft.

Buginga, G. and de Souza e Silva, E. (2024). Clustering survival data using a mixture of non-parametric experts. arXiv.

Buginga, G. C. (2023). Algoritmos de Aprendizado de Máquina Aplicados a Dados Censoriados para Previsão de Mortalidade de Pacientes com Doença Arterial Coronariana. Dissertação de mestrado, COPPE - Universidade Federal do Rio de Janeiro.

de Almeida, C. M., Leão, R. M. M., and de Souza e Silva, E. (2024). Inferindo pontos de mudança em séries temporais com dados não rotulados: um breve estudo usando dados do NDT. In XLII SBRC, pages 686–699.

de Souza e Silva, C. G., Buginga, G. C., de Souza e Silva, E. A., Arena, R., Rouleau, C. R., Aggarwal, S., Wilton, S. B., Austford, L., Hauer, T., and Myers, J. (2022). Prediction of mortality in coronary artery disease: Role of machine learning and maximal exercise capacity. Mayo Clinic Proceedings, 97(8):1472–1482.

de Souza e Silva, E. and Gail, H. R. (1989). Calculating availability and performability measures of repairable computer systems using randomization. J. ACM, 36(1):171–193.

Gill, P., Diot, C., Ohlsen, L. Y., Mathis, M., and Soltesz, S. (2022). M-lab: user initiated internet data for the research community. SIGCOMM Comput. Commun. Rev., 52(1):34–37.

M-Lab (2025). Ndt (network diagnostic tool).

Pi, Y., Jamin, S., and Wei, L. (2024). Measuring congestion-induced performance imbalance in internet load balancing at scale. In Proceedings of the ACM SIGCOMM.

RNP (2023). RNP e M-Lab assinam MoU. [link].

Sharma, A., Sundaresan, S., and Madhyastha, H. V. (2024a). Measuring the prevalence of wifi bottlenecks in home access networks. In Proceedings of the ACM Internet Measurement Conference (IMC).

Sharma, T., Padaki, R., and Feamster, N. (2024b). Beyond data points: Regionalizing crowdsourced latency measurements. In Proceedings of the ACM SIGMETRICS. ACM.

Streit, A. G., dos Santos, G. H. A., Leão, R. M. M., e Silva, E. d. S., Menasché, D., and Towsley, D. (2021). Network anomaly detection based on tensor decomposition. Computer Networks, 200:108503.