Escalability Evaluation of High-Performing Applications in Public and Private Clouds
Abstract: The analysis of high-performance computing application (HPC), has recently been facilitated by the use of cloud platforms. However, the performance of HPC applications relies heavily on the support I/O platform, mainly the communication network between the VMs. In this study, the performance of the NPB-NAS Parallel Benchmark suite and DGEO suite, which processes spatial queries, were compared in a physical cluster, a private cloud (XenServer), and a public cloud (Microsoft Azure). The experiment demonstrates a speed-up up to 3.0 times lower in the public cloud for CPU-bounded applications, and a runtime up to 3.0 times greater compared to the physical cluster. The speed-up between cloud platforms tested here is not affected only by the extra layer of virtualization, but also by other factors such as VM neighbors.
Keywords: .
Citation: Ariel Lauber de Paula Silva; Thiago Borges de Oliveira. Avaliação da Escalabilidade de Aplicações de Alto Desempenho em Nuvem Pública e Privada. In: Anais do WSCAD 2016, Aracajú, SE, Brasil, 2016, pp. 1--12.
Copy citation in bibtex format.