Modeling and categorization of user trajectories in supercomputing systems using multi-aspect graphs

• Arthur Henrique Craveiro Costa LNCC
• Juan Humberto Leonardo Fabian LNCC
• Antônio Tadeu Azevedo Gomes LNCC

DOI: https://doi.org/10.5753/sscad.2026.34965

Resumo

Efficient resource management in supercomputing systems requires an accurate characterization of user behavior. Traditional workload characterization methods often fail to capture the multidimensional interactions between users and these systems. In this work, we propose a novel framework for modeling and categorizing user trajectories in supercomputers using Multi-Aspect Graphs (MAG). In our approach, each job is represented as a MAG vertex integrating five complementary aspects: requested queue, quantized CPU and memory utilization, user context, chronological submission order, and computational magnitude. Directed edges connect consecutive jobs, enriched with user identifiers and calculated “think times” to maintain authorship and temporal dynamics. To compare user behaviors, we adapt a sequence alignment algorithm for MAGs, combining aspect and temporal dissimilarity metrics. Subsequent trajectory clustering enables the identification of distinct, overlapping behavioral profiles. We evaluated our framework on a real-world job log, demonstrating its ability to preserve individual trajectory structures, reveal submission cadences, and group users with similar behavioral patterns without collapsing distinct dimensions into generic statistical aggregates.

Referências

Bose, A., Yang, H., Hsu, W. H., and Andresen, D. (2021). HPCGCN: A predictive framework on high performance computing cluster log data using graph convolutional networks. In 2021 IEEE International Conference on Big Data, pages 4113–4118. IEEE.

Caliński, T. and Harabasz, J. (1974). A dendrite method for cluster analysis. Communications in Statistics, 3(1):1–27.

Collins, L. M. and Dent, C. W. (1988). Omega: A general formulation of the rand index of cluster recovery suitable for non-disjoint solutions. Multivariate Behavioral Research, 23(2):231–242.

Davies, D. L. and Bouldin, D. W. (1979). A cluster separation measure. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-1(2):224–227.

dos Santos, J. P. M. N., da Silva, J. E. H., and Gomes, A. T. A. (2026). Towards workload-tailored optimization of job scheduling policies in HPC environments. Concurrency and Computation: Practice and Experience, 38(1):e70532.

dos Santos, J. P. M. N. and Gomes, A. T. A. (2024). Impact of job scheduling policy changes on user behaviour and system response: The case of the Santos Dumont supercomputer in Brazil. In Latin American High Performance Computing Conference. Springer.

Feitelson, D. G. (2001). Metrics for parallel job scheduling and their convergence. In Feitelson, D. G. and Rudolph, L., editors, Job Scheduling Strategies for Parallel Processing, pages 188–205, Berlin, Heidelberg. Springer Berlin Heidelberg.

Jakobsche, T., Lachiche, N., and Ciorba, F. M. (2023). Investigating HPC job resource requests and job efficiency reporting. In 2023 22nd International Symposium on Parallel and Distributed Computing, pages 61–68. IEEE.

Lancichinetti, A., Radicchi, F., Ramasco, J. J., and Fortunato, S. (2011). Finding statistically significant communities in networks. PLOS ONE, 6(4):e18961.

Lim, S.-H., Sim, H., Gunasekaran, R., and Vazhkudai, S. S. (2017). Scientific user behavior and data-sharing trends in a petascale file system. In Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis. ACM.

MacQueen, J. (1967). Some methods for classification and analysis of multivariate observations. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, volume 1, pages 281–297, Berkeley, CA. University of California Press.

Madon, M., Costa, G. D., and Pierson, J.-M. (2024). Replay with feedback: How does the performance of HPC system impact user submission behavior? Future Generation Computer Systems, 155:66–79.

Needleman, S. B. and Wunsch, C. D. (1970). A general method applicable to the search for similarities in the amino acid sequence of two proteins. Journal of Molecular Biology, 48(3):443–453.

Patel, T., Rich, P., Allcock, W., Liu, Z., Kettimuthu, R., and Tiwari, D. (2020). Job characteristics on large-scale systems: Long-term analysis, quantification, and implications. In SC20: International Conference for High Performance Computing, Networking, Storage and Analysis. IEEE.

Rosa, C. O. C. S., Ito, M., Vieira, A. B., Wehmuth, K., and Gomes, A. T. A. (2023). Framework based on complex networks to model and mine patient pathways. arXiv preprint arXiv:2309.14208. arXiv:2309.14208v2 [cs.CY].

Rousseeuw, P. J. (1987). Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20:53–65.

SchedMD (2023). Slurm Workload Manager Documentation. SchedMD LLC.

Schlagkamp, S., Ferreira da Silva, R., Allcock, W., Deelman, E., and Schwiegelshohn, U. (2016). Consecutive job submission behavior at Mira supercomputer. In Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing, HPDC ’16, page 93–96, New York, NY, USA. Association for Computing Machinery.

Simakov, N. A., White, J. P., DeLeon, R. L., Gallo, S. M., Jones, M. D., Palmer, J. T., Plessinger, B. D., and Furlani, T. R. (2018). A workload analysis of NSF’s innovative HPC resources using XDMoD. CoRR, abs/1801.04306.

Wang, Q., Shen, Y., and Li, J. (2021). User-level workload analysis for supercomputers. In Proceedings of the 4th International Conference on Software Engineering and Information Management, pages 68–73. ACM.

Ward, Jr., J. H. (1963). Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association, 58(301):236–244.

Wehmuth, K., Fleury, É., and Ziviani, A. (2016). On MultiAspect graphs. Theor. Comput. Sci., 651:50–61.

You, H. and Zhang, H. (2013). Comprehensive workload analysis and modeling of a petascale supercomputer. In Job Scheduling Strategies for Parallel Processing, volume 7698 of Lecture Notes in Computer Science, pages 253–271. Springer.

Zheng, S., Shae, Z.-Y., Zhang, X., Jamjoom, H., and Fong, L. (2011). Analysis and modeling of social influence in high performance computing workloads. In Euro-Par 2011 Parallel Processing, volume 6852 of Lecture Notes in Computer Science, pages 193–204. Springer.