Performance analysis of Cloud Computing storage services for checkpoint execution

• Charles B. Rodamilans Mackenzie
• Edson Borin UNICAMP https://orcid.org/0000-0003-1783-4231

DOI: https://doi.org/10.5753/eradsp.2020.16893

Abstract

Cloud computing has been used to execute high-performance computing applications due to its potential to reduce costs. The instability of low-cost instances - Spots - can be mitigated with checkpointing, and the performance of the storage service becomes crucial to avoid increasing the total runtime excessively. We compared the performance of four Cloud Computing storages services on AWS (EBS, EFS, FSx for Lustre e S3) for checkpoint persistence and observed that: (1) two of the four storage services showed scalability; (2) the storage service can increase the write performance by up to 727%.

Keywords: Cloud Computing, High Performance Computing, Spot, Storage

References

Cao, J., Kerr, G., Arya, K., and Cooperman, G. (2014). Transparent checkpoint-restart over infiniband. In Proceedings of the 23rd international symposium on High- performance parallel and distributed computing, pages 13–24.

Netto, M. A., Calheiros, R. N., Rodrigues, E. R., Cunha, R. L., and Buyya, R. (2018). Hpc cloud for scientific and business applications: Taxonomy, vision, and research challenges. ACM Computing Surveys (CSUR), 51(1):1–29.