Avaliação de Lustre e GekkoFS sob Cinco Estratégias de I/O em uma Aplicação CFD
Resumo
Sistemas de Computação de Alto Desempenho (HPC) dependem de sistemas de arquivos paralelos capazes de sustentar alto throughput e baixa latência. Cargas de trabalho modernas, caracterizadas por operações intensivas de metadados e alta concorrência, desafiam sistemas tradicionais como o Lustre, cujo servidor de metadados centralizado constitui um gargalo estrutural. O GekkoFS responde a essa limitação distribuindo metadados e dados nos nós de computação. Este artigo avalia ambos os sistemas em dois cenários. Com os benchmarks MDTest e IOR, o GekkoFS supera o Lustre em operações de metadados por até uma ordem de magnitude, enquanto o Lustre permanece competitivo em throughput sequencial. Com uma aplicação de Dinâmica de Fluidos Computacional (CFD) multi-GPU, cinco padrões de escrita são comparados. No Lustre, a escrita assíncrona com múltiplos arquivos apresenta o menor overhead (6,6%), ao passo que a consolidação em arquivo único apresenta o maior (53,7%), evidenciando o deslocamento do gargalo do servidor de metadados para o servidor de dados. No GekkoFS, os modos assíncronos com arquivo único são os mais eficientes (12,0% e 12,3%).
Referências
Astsatryan, H., Dovrolis, C., and Baghdasaryan, H. (2026). Convergence of HPC and cloud storage systems for deep learning workflows: An evaluation of LustreFS and S3-compatible object storage. IEEE Access.
Balouek, D., Carpen Amarie, A., Charrier, G., Desprez, F., Jeannot, E., Jeanvoine, E., Lèbre, A., Margery, D., Niclausse, N., Nussbaum, L., Richard, O., Pérez, C., Quesnel, F., Rohr, C., and Sarzyniec, L. (2013). Adding virtualization capabilities to the Grid’5000 testbed. In Cloud Computing and Services Science, volume 367 of Communications in Computer and Information Science, pages 3–20. Springer.
Carneiro, A. R., Bez, J. L., Boito, F. Z., Fagundes, B. A., Osthoff, C., and Navaux, P. O. A. (2023). Uncovering I/O demands on HPC platforms: Peeking under the hood of Santos Dumont. Journal of Parallel and Distributed Computing, 182:104744.
Cluster File Systems, Inc. (2007). Lustre: A scalable, high-performance file system. White paper, Cluster File Systems, Inc.
Depardon, B., Le Mahec, G., and Séguin, C. (2013). Analysis of six distributed file systems. Technical Report hal-00789086, HAL.
Folk, M., Heber, G., Koziol, Q., Pourmal, E., and Robinson, D. (2011). An overview of the HDF5 technology suite and its applications. In Proceedings of the EDBT/ICDT 2011 Workshop on Array Databases, AD ’11, pages 36–47, New York, NY, USA. Association for Computing Machinery.
Godoy, W. F., Podhorszki, N., Wang, R., Atkins, C., Eisenhauer, G., Gu, J., Davis, P., Choi, J., Germaschewski, K., Huck, K., Huebl, A., Kim, M., Kress, J., Kurc, T., Liu, Q., Logan, J., Mehta, K., Ostrouchov, G., Parashar, M., Poeschel, F., Pugmire, D., Suchyta, E., Takahashi, K., Thompson, N., Tsutsumi, S., Wan, L., Wolf, M., Wu, K., and Klasky, S. (2020). ADIOS 2: The adaptable input output system. a framework for high-performance data management. SoftwareX, 12:100561.
Grid’5000 (2026). Grid’5000: Home. [link]. Acesso em: 19 jun. 2026.
IOR Project (s.d.). IOR and mdtest: Parallel I/O and metadata benchmarks. [link]. Repositório oficial. Acesso em: 19 out. 2026.
Laitala, J. (2017). Metadata management in distributed file systems. Master’s thesis, University of Helsinki.
Liang, Z., Lombardi, J., Chaarawi, M., and Hennecke, M. (2020). DAOS: A scale-out high performance storage stack for storage class memory. In Supercomputing Frontiers, volume 12082 of Lecture Notes in Computer Science, pages 40–54, Cham. Springer.
Lockwood, G. K., Snyder, S., Wang, T., Byna, S., Carns, P., and Wright, N. J. (2018). A year in the life of a parallel file system. In SC18: International Conference for High Performance Computing, Networking, Storage and Analysis, pages 931–943. IEEE.
Luu, H., Winslett, M., Gropp, W., Ross, R., Carns, P., Harms, K., Prabhat, M., Byna, S., and Yao, Y. (2015). A multiplatform study of I/O behavior on petascale supercomputers. In Proceedings of the 24th International Symposium on High-Performance Parallel and Distributed Computing (HPDC), pages 33–44.
Ravi, J., Byna, S., Koziol, Q., Tang, H., and Becchi, M. (2023). Evaluating asynchronous parallel I/O on HPC systems. In 2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS), pages 211–221. IEEE.
Tarraf, A., Bandet, A., Boito, F., Pallez, G., and Wolf, F. (2024). I/O behind the scenes: Bandwidth requirements of HPC applications with asynchronous I/O. In 2024 IEEE International Conference on Cluster Computing (CLUSTER), pages 426–439. IEEE.
ThinkParQ and Fraunhofer ITWM (2024). BeeOND: BeeGFS On Demand. ThinkParQ. Documentação do BeeGFS, versão 7.2.13. Acesso em: 19 out. 2026.
Vef, M.-A., Moti, N., Süß, T., Tacke, M., Tocci, T., Nou, R., Miranda, A., Cortes, T., and Brinkmann, A. (2020). GekkoFS: A temporary burst buffer file system for HPC applications. Journal of Computer Science and Technology, 35:72–91.
Vef, M.-A., Moti, N., Süß, T., Tocci, T., Nou, R., Miranda, A., Cortes, T., and Brinkmann, A. (2018). GekkoFS: A temporary distributed file system for HPC applications. In 2018 IEEE International Conference on Cluster Computing (CLUSTER), pages 319–324. IEEE.
Wang, Y., Chen, B., Chen, C., Zhu, J., Xiang, B., Heaney, C. E., and Pain, C. C. (2024). Artificial intelligence for partial differential equations in computational mechanics: A review. Applied Mechanics Reviews, pages 1–81.
Balouek, D., Carpen Amarie, A., Charrier, G., Desprez, F., Jeannot, E., Jeanvoine, E., Lèbre, A., Margery, D., Niclausse, N., Nussbaum, L., Richard, O., Pérez, C., Quesnel, F., Rohr, C., and Sarzyniec, L. (2013). Adding virtualization capabilities to the Grid’5000 testbed. In Cloud Computing and Services Science, volume 367 of Communications in Computer and Information Science, pages 3–20. Springer.
Carneiro, A. R., Bez, J. L., Boito, F. Z., Fagundes, B. A., Osthoff, C., and Navaux, P. O. A. (2023). Uncovering I/O demands on HPC platforms: Peeking under the hood of Santos Dumont. Journal of Parallel and Distributed Computing, 182:104744.
Cluster File Systems, Inc. (2007). Lustre: A scalable, high-performance file system. White paper, Cluster File Systems, Inc.
Depardon, B., Le Mahec, G., and Séguin, C. (2013). Analysis of six distributed file systems. Technical Report hal-00789086, HAL.
Folk, M., Heber, G., Koziol, Q., Pourmal, E., and Robinson, D. (2011). An overview of the HDF5 technology suite and its applications. In Proceedings of the EDBT/ICDT 2011 Workshop on Array Databases, AD ’11, pages 36–47, New York, NY, USA. Association for Computing Machinery.
Godoy, W. F., Podhorszki, N., Wang, R., Atkins, C., Eisenhauer, G., Gu, J., Davis, P., Choi, J., Germaschewski, K., Huck, K., Huebl, A., Kim, M., Kress, J., Kurc, T., Liu, Q., Logan, J., Mehta, K., Ostrouchov, G., Parashar, M., Poeschel, F., Pugmire, D., Suchyta, E., Takahashi, K., Thompson, N., Tsutsumi, S., Wan, L., Wolf, M., Wu, K., and Klasky, S. (2020). ADIOS 2: The adaptable input output system. a framework for high-performance data management. SoftwareX, 12:100561.
Grid’5000 (2026). Grid’5000: Home. [link]. Acesso em: 19 jun. 2026.
IOR Project (s.d.). IOR and mdtest: Parallel I/O and metadata benchmarks. [link]. Repositório oficial. Acesso em: 19 out. 2026.
Laitala, J. (2017). Metadata management in distributed file systems. Master’s thesis, University of Helsinki.
Liang, Z., Lombardi, J., Chaarawi, M., and Hennecke, M. (2020). DAOS: A scale-out high performance storage stack for storage class memory. In Supercomputing Frontiers, volume 12082 of Lecture Notes in Computer Science, pages 40–54, Cham. Springer.
Lockwood, G. K., Snyder, S., Wang, T., Byna, S., Carns, P., and Wright, N. J. (2018). A year in the life of a parallel file system. In SC18: International Conference for High Performance Computing, Networking, Storage and Analysis, pages 931–943. IEEE.
Luu, H., Winslett, M., Gropp, W., Ross, R., Carns, P., Harms, K., Prabhat, M., Byna, S., and Yao, Y. (2015). A multiplatform study of I/O behavior on petascale supercomputers. In Proceedings of the 24th International Symposium on High-Performance Parallel and Distributed Computing (HPDC), pages 33–44.
Ravi, J., Byna, S., Koziol, Q., Tang, H., and Becchi, M. (2023). Evaluating asynchronous parallel I/O on HPC systems. In 2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS), pages 211–221. IEEE.
Tarraf, A., Bandet, A., Boito, F., Pallez, G., and Wolf, F. (2024). I/O behind the scenes: Bandwidth requirements of HPC applications with asynchronous I/O. In 2024 IEEE International Conference on Cluster Computing (CLUSTER), pages 426–439. IEEE.
ThinkParQ and Fraunhofer ITWM (2024). BeeOND: BeeGFS On Demand. ThinkParQ. Documentação do BeeGFS, versão 7.2.13. Acesso em: 19 out. 2026.
Vef, M.-A., Moti, N., Süß, T., Tacke, M., Tocci, T., Nou, R., Miranda, A., Cortes, T., and Brinkmann, A. (2020). GekkoFS: A temporary burst buffer file system for HPC applications. Journal of Computer Science and Technology, 35:72–91.
Vef, M.-A., Moti, N., Süß, T., Tocci, T., Nou, R., Miranda, A., Cortes, T., and Brinkmann, A. (2018). GekkoFS: A temporary distributed file system for HPC applications. In 2018 IEEE International Conference on Cluster Computing (CLUSTER), pages 319–324. IEEE.
Wang, Y., Chen, B., Chen, C., Zhu, J., Xiang, B., Heaney, C. E., and Pain, C. C. (2024). Artificial intelligence for partial differential equations in computational mechanics: A review. Applied Mechanics Reviews, pages 1–81.
Publicado
03/11/2026
Como Citar
VARGAS, João V.; KÜNAS, Cristiano A.; NAVAUX, Philippe O. A.; MORALES, Bruno M..
Avaliação de Lustre e GekkoFS sob Cinco Estratégias de I/O em uma Aplicação CFD. In: SIMPÓSIO EM SISTEMAS COMPUTACIONAIS DE ALTO DESEMPENHO (SSCAD), 27. , 2026, Natal/RN.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 538-551.
DOI: https://doi.org/10.5753/sscad.2026.34981.
