Previsão da Classe de Frequência de Acesso de Objetos em Serviços de Armazenamento em Nuvem

• Flávio A. A. Motta UFV
• Patrick R. P. Lemes UFJF
• Glauber D. Golçalves UFPI
• Heder S. Bernardino UFJF
• Saulo M. Villela UFJF
• Alex B. Vieira UFV / UFJF

DOI: https://doi.org/10.5753/sbrc.2022.222338

Abstract

Cloud storage services offer to domestic and corporate users advantages as backup, data replication in different locations, data sharing, and collaborative work. Additionally, providers of these services offer tiered cloud storage with multiple pricing options based on the level of storage used. In this work, we investigate a relevant aspect regarding costs of this service for users: predicting the data access class as frequent or infrequent and allocating it in a suitable storage tier. In this sense, we propose a machine learning framework that predicts the appropriate classes based on data access patterns. We evaluated the performance of this model through data trace-oriented of a real service. Data allocation methods used in the literature demonstrate an improvement potential of up to 41% over traditional cloud storage methods. Our results show that there is a potential for storage cost savings of up to 15.92% when compared to data allocation methods used in the literature.

References

Bocchi, E., Drago, I., and Mellia, M. (2015). Personal Cloud Storage: Usage, Performance and Impact of Terminals. In Proc. of the IEEE CloudNet.

Chawla, N. V., Bowyer, K. W., Hall, L. O., and Kegelmeyer, W. P. (2002). Smote: Synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 16:321–357.

Cisco (2019). Cisco Global Cloud Index: Forecast and Methodology, 2016–2021 White Paper. Disponível em https://www.cisco.com Document ID 1513879861264127.

Erradi, A. and Mansouri, Y. (2020). Online cost optimization algorithms for tiered cloud storage services. Journal of Systems and Software, 160:110457.

Gonçalves, G., Drago, I., Silva, A. P. C., Vieira, A. B., and Almeida, J. M. (2016). The IEEE Internet impact of content sharing on cloud storage bandwidth consumption. Computing, 20(4):26–35.

Gracia-Tinedo, R., García-López, P., Gómez, A., and Illana, A. (2016). Understanding data sharing in private personal clouds. In Proc. of the IEEE CLOUD.

Hsu, Y., Irie, R., Murata, S., and Matsuoka, M. (2018). A novel automated cloud storage tiering system through hot-cold data classification. In Proc. of the IEEE CLOUD.

Irie, R., Murata, S., Hsu, Y., and Matsuoka, M. (2018). A novel automated tiered storage architecture for achieving both cost saving and qoe. In Proc. of the IEEE SC2.

Kaushik, R. T. and Bhandarkar, M. (2010). Greenhdfs: towards an energy-conserving, storage-efficient, hybrid hadoop compute cluster. In Proc. of the USENIX.

Liu, M., Pan, L., and Liu, S. (2019). To transfer or not: An online cost optimization algorithm for using two-tier storage-as-a-service clouds. IEEE Access, 7:94263–94275.

Liu, M., Pan, L., and Liu, S. (2021). Keep hot or go cold: A randomized online migration algorithm for cost optimization in staas clouds. IEEE Transactions on Network and Service Management.

Mesnier, M., Ganger, G. R., and Riedel, E. (2003). Object-based storage. IEEE Communications Magazine, 41(8):84–90.

Ribeiro, S., Gonçalves, G., Silva, F., Vieira, A., and Almeida, J. (2020). Análise de um serviço virtual de armazenamento que explora classes de objetos na nuvem e padrões de acesso. In Anais do XIX Workshop em Desempenho de Sistemas Computacionais e de Comunicação, pages 85–96, Porto Alegre, RS, Brasil. SBC.

Subramanian, M., Lloyd, W., Roy, S., Hill, C., Lin, E., Liu, W., Pan, S., Shankar, S., Viswanathan, S., Tang, L., and Kumar, S. (2014). f4: Facebook’s warm blob storage system. In Proc. of the OSDI.